Engaged by the Initiative? How the Use of Citizen Initiatives Increases Voter Turnout

Abstract

Using data from 1870 to 2008, the authors attempt to resolve competing claims about the nature of how citizen initiatives affect turnout in the American states. They provide evidence that mobilization is the mechanism through which direct democracy increases turnout. Contrary to previous research, they show that the adoption of the initiative and past usage of the process do not lead to higher turnout in a given election. Citizen initiative campaigns mobilize the electorate in current elections, and the number of competitive initiative elections has a greater effect on turnout than uncompetitive races.

Keywords

elections voting behavior public opinion political participation state politics policy

Research on political behavior has focused on a range of causal factors that increase political participation, voter turnout specifically. The discipline has examined external influences such as changes in electoral institutions designed to negatively affect voter turnout, poll taxes (Key 1949/1984; J. M. Kousser 1974), or to increase turnout with mail ballots (Karp and Banducci 2000; T. Kousser and Mullin 2007) and has examined internal influences such as genetics (Fowler and Dawes 2008; Fowler, Baker, and Dawes 2008). Aside from the direct intended effects of institutional changes, some American political institutions have had unintended consequences for political participation, as in the imposition of the secret ballot and the ensuing decrease in voter turnout (Heckelman 1995). We argue that direct democracy has had the opposite unintended effect, as campaigns for initiatives, particularly competitive races, mobilize voters and increase turnout. Our results show that the existence of the institution of direct democracy in and of itself does not lead to increases in voter turnout. However, our evidence suggests that the campaigns associated with initiatives do increase turnout.

In this article, we make three fundamental contributions to the literature on direct democracy and political participation that clarify how ballot propositions engage the electorate. First, we directly test competing theories of how citizen initiatives affect turnout by measuring the effects of the introduction and usage of the institution. Second, we investigate the differential effects that initiatives have on turnout by using competitiveness as a proxy for campaign intensity. Third, our historical data set allows us to draw proper causal inferences about the relationship between citizen initiatives and voter turnout.

Our work shows that merely having the initiative process in the state or simply having used it in the past does not affect turnout in a given election. Compared to states without the initiative process, states with initiatives on the ballot have higher turnout in that election. As the number of initiatives on a ballot rises during midterm elections, voter turnout does as well, but these effects do so with diminishing marginal returns as the number of initiatives increases. Controlling for changes in demographics over time (percentage foreign born, percentage nonwhite, and urbanization), institutional changes (adoption of the secret ballot and expanding suffrage to women), contextual electoral effects (party competition and the presence of gubernatorial and senate elections), and proposition-specific characteristics (competitiveness), we show that citizen initiatives (both competitive and uncompetitive—though less so) positively affect voter turnout during midterm elections and wield no appreciable influence on turnout in presidential elections.

The results presented here suggest that the causal story of why direct democracy leads to an increase in turnout is one of mobilization. We contend that political campaigns, and their ensuing mobilization effects, are the mechanisms that lead to the increases in turnout associated with direct democracy. In the following sections we review previous research on initiatives, explain our hypotheses and methodology, discuss our results, and conclude with potential directions for future research.

Direct Democracy and Turnout

Scholars concerned with the health of the U.S. democracy have decried Americans’ relatively declining levels of political engagement in the past few decades, especially regarding voter turnout (Patterson 2002; Putnam 1995). However, if one measures voter turnout using the voting-eligible population (VEP) instead of the voting-age population (VAP), one will see that voter turnout in recent elections was as high or higher than it was in the 1948 presidential election (McDonald and Popkin 2001). Despite Putnam’s (1995) and Patterson’s (2002) claims, recent studies have shown that Americans today are generally more interested in elections and are more familiar with presidential candidates’ and parties’ policy positions than they were in the 1970s (Childers and Popkin 2007). Ballot propositions can set the political agenda (Nicholson 2005) and can also serve as one factor that can increase turnout in elections, especially in the absence of major sources of campaign mobilization (Schlozman and Yohai 2008; D. A. Smith and Tolbert 2004; Tolbert, Bowen, and Donovan 2009; Tolbert, Grummel, and Smith 2001).

It has been argued that there are two components to the turnout bump associated with initiatives: election-specific effects and institutional effects. Studies analyzing election-specific effects have provided evidence that campaign intensity surrounding the initiative races provides voters with information, which lowers the costs of their electoral decision-making process and in turn makes them more likely to vote (D. A. Smith and Tolbert 2004; Tolbert and Smith 2005; Tolbert, Bowen, and Donovan 2009; Tolbert, Grummel, and Smith 2001). While initiatives are typically low-information affairs (Lupia 1994; Magleby 1984), highly competitive races, such as California’s Proposition 8 in 2008, can generate intensely information-rich campaigns.

Studies in the election-specific effects camp have generally argued that while initiatives have specific policy purposes, they also tend to have a series of unintended positive effects on people’s levels of political engagement. Summarizing recent research, Tolbert and Smith (2005) argue that initiatives create an atmosphere with a more engaged citizenry. The presence of ballot initiatives leads to a better informed public, one that is more confident in government, and therefore they more likely to vote. They use the number of initiatives as a proxy for campaign effects to show that as the number of initiatives on the ballot increases, turnout does as well in both presidential and midterm elections.

The “educated by initiative” theory leaves out an important part of the causal argument about how initiatives engage the electorate: campaigns. While some ballot propositions involve issues with enough specific interest to draw some voters to the polls, there is nothing inherently distinctive about initiatives in general that cause people to vote. Campaigns provide the public with information about the issues and highlight their significance to voters in the interest of getting them out to vote. That is why the most intense elections garner the highest turnout rates (Goldstein and Freedman 2002).

Aside from direct election-specific mobilization effects, it has been argued that there are also institutional effects at work in direct democracy. We consider institutional effects to be any effects emanating from having adopted the initiative process but not having initiatives on the ballot in the current election. These institutional effects would affect turnout over multiple election years, and they may include having used the initiative process in the past (regardless of how often) or just having the process itself. Hero and Tolbert (2004) find that the average number of initiatives over the previous twenty years has a positive effect on turnout. Bowler and Donovan (2002) find that the cumulative number of initiatives a state has used since adopting the process increases turnout in a given election. Tolbert, Grummel, and Smith (2001) show that having the initiative process itself significantly increases turnout independent of the number of initiatives a state has on the ballot in midterm elections, though at the individual level using voter lists, Dyck and Seabrook (2010) present evidence that shows there is minimal, if any, long-term influence on voter turnout from having lived in states with the initiative process.

Such studies imply that initiatives may draw marginal voters to the polls, but these people are then more likely to vote in subsequent elections. This is congruent with experimental and observational research on voting behavior showing voting is habitual: once you vote in one election, you are more likely to vote in subsequent ones (Gerber, Green, and Shachar 2003). Schlozman and Yohai (2008) test alternative measures of the effects of historical initiative usage and find that in most midterm election years since 1978, and in the 1992 and 1996 presidential elections, turnout is not consistently related to the number of initiatives a state has had on the ballot in previous years. They conclude that there is some minor carryover from previous elections, but the effects of initiatives are specific to a particular electoral cycle, as they do not observe reliable results across the election years in their study.

Scholars, however, should proceed with caution in taking guidance from prior studies about initiative usage and turnout. The works discussed above begin their analysis of the effects initiatives can have on turnout using elections beginning in the 1970s, decades after most initiative states adopted the institution into their constitutions. Because of the time frame used in those studies, they could not test the effect that the adoption of the process has had on voter engagement. In addition, most (though not all) of these studies ignore the possibility that states that adopted the initiative may already have had a relatively more politically engaged electorate than states that did not.

Tolbert, Bowen, and Donovan (2009) do provide a direct measure of campaign effects by using campaign spending data in only those states with the initiative process. This design allows them to partially mitigate the endogeneity problem. Unfortunately, they are able to study only a few elections since 2000 because spending data for initiatives are not widely available for campaigns prior to 2000. Of the three elections they look at, they find evidence that campaign spending affects turnout only in 2004 and 2006. Even with the limited time frame and uneven results, these results suggest that campaign mobilization is the driving force for increasing turnout. Nonetheless, several questions remain unanswered about how citizen initiatives affect turnout. Did the adoption of the initiative permanently increase turnout in states that adopted the process? Do initiatives exhibit election-specific effects (affecting turnout in a given election), or do initiatives in one election boost turnout in subsequent elections, exhibiting a permanent effect? Do all initiatives have equal effects on turnout, or do relatively competitive elections affect turnout more than others?

In the following sections we examine the current claims about initiatives using a research design that gives us more leverage on estimating causal relationships (election specific and institutional effects) between initiatives and turnout. We employ a hybrid time-series quasi-experiment to isolate the effects of initiative usage on turnout. This analysis uses electoral competition as a proxy for mobilization and provides compelling evidence that the presence of initiative campaigns can induce the marginal voter to show up at the polls in midterm elections. We also show that just having the initiative process does not permanently increase turnout in the states with direct democracy, and past initiative usage does not significantly affect turnout in current elections.

Theory

We argue that if initiatives stimulate voter turnout, they do so through campaign mobilization. Campaigns lower people’s costs of voting by providing them with information about candidates and issues, work to convince people their candidate (or their position on the proposition) is closest to them on the issues, and do everything they can to get their supporters to the polls on Election Day. Most people are casual observers of politics (Downs 1957), and political competition is a necessary condition to get them to tune in to politics, and campaigns drive them to the polls, particularly in initiative races (Dyck and Seabrook 2010).

While the specific techniques available to campaigns have evolved over time because of advances in technology, the fundamentals of campaign strategies are largely the same today as they were in the nineteenth century. Campaigns spend more and more resources on outreach and mobilization efforts as a function of how close they expect the election to be. This is certainly true in presidential elections (Shaw 2006), but it also occurs in others down the ballot (Jacobson 2004). With polling technologies being what they are today and a flourishing industry of political consultants from both political parties (including a robust permanent initiative campaign industry; see Magleby 1984), campaigns at all levels are able to read the political tea leaves early on in the electoral cycle. However, even in the nineteenth century, party officials used their organizations to poll local districts and even entire states and were able to gauge how close an election would be. This allowed them to allocate their resources to the areas where elections were expected to be closest (Marcus 1971, 12–13).

Politicians could usually tell how various groups would vote. They talked, as politicians usually do, about the “Germans,” or the “wheat farmers” or the “wool-growing counties.” They knew roughly where there were Democrats and where Republicans. They could predict with considerable accuracy sure victories, certain defeats, and which contests would be close and require attention. . . . Politicians calculated their chances by handfuls of votes. They pinpointed the difference, and sought means to gain them. Their principle [sic] method was to stir up “enthusiasm” to bring their potential voters to the polls. (Marcus 1971, 11–13)

Campaigns in the late nineteenth century and early twentieth century reallocated their resources from hopeless races to more competitive ones. Some of their methods included purchasing votes, letting people vote repeatedly, stuffing ballot boxes, falsifying returns, and even bringing in people from out of the district (Marcus 1971). For example, Democrats feared Republicans were moving African Americans from the South to the more competitive northern states. “Our people are becoming alarmed at a trainload of negroes who are passing through here going from the South to northern states. . . . We fear that these negroes are being taken to some of the large cities where they will register their vote against our party” (Blair 1916).

While we argue that initiatives affect turnout via campaign mobilization, we do not expect them to matter equally. In presidential election years, coverage of the presidential campaigns drowns out down-ballot elections and makes it difficult for issue elections to gain traction among voters. Schlozman and Yohai (2008) find that initiatives do not consistently affect turnout during presidential years. Even highly salient, controversial initiatives dealing with same-sex marriage did not have any appreciable effects on turnout in the 2004 election (Abramowitz 2004; D. A. Smith, DeSantis, and Kassel 2006). Initiative campaigns are thus more likely to successfully connect with voters during off-year elections when they do not have to compete with presidential campaigns. Nicholson (2003) shows that voters are more aware of ballot propositions during midterm elections because they are more likely to get media coverage as state issues are more likely to be salient (compared to presidential election years).

Hypotheses and Research Design

Our expectation, based on consistent findings in the literature, is that if ballot propositions have any effect on turnout, it will be greatest during lower information midterm elections. Our theory predicts that competitive races engender greater mobilization efforts and will in turn increase turnout more so than uncompetitive races.

The theory of mobilization driving turnout predicts that the mere adoption of the initiative process will not have any effect on turnout. Most people are not strongly interested in politics, and they focus their attention on what matters most in getting through their daily lives (Downs 1957; Popkin 1994). Consequently we are skeptical that any effects from a relatively intense atmosphere via initiative campaigns would continue to affect election years into the future. After the election is over, most people who are not political junkies will default back to their typical levels of political interest. Ergo, we also do not expect to find that the previous usage of initiatives will affect turnout in current elections, meaning states without initiatives on the ballot will not experience increases in turnout.

Initiatives find their way to the ballot without going through the legislature and often involve narrow moneyed interests (Ernst 2001) or very salient issues (M. A. Smith 2001). Even the most uninteresting ballot initiative has to go through the signature-gathering phase, a nontrivial expense, especially in large states such as California. Therefore, we would expect uncompetitive initiative races to engender some campaign activity, though not nearly as much as competitive initiative races. Below we formally state our hypotheses derived from the logic of the theory of institutional effects and our theory of mobilization.

Hypotheses (Institutional Effects):

Hypothesis 1 (H₁): Adopting the citizen initiative process will lead to an increase in statewide turnout compared to that in noninitiative states.

Hypothesis 2 (H₂): Initiative states will have higher turnout than noninitiative states after the process has been used, even when initiatives are not on the ballot.

Hypotheses (Election-Specific Effects):

Hypothesis 3 (H₃): States with initiatives on the ballot will have higher turnout than states without initiatives on the ballot.

Hypothesis 4 (H₄): The number of competitive initiatives on the ballot will have a greater effect on turnout than the number of uncompetitive initiatives on the ballot.

Data and Method

In studies where the treatment effect is not randomly assigned, researchers face a number of confounding factors that can bias estimates of treatment effects. This is especially the case when studies fail to measure pretreatment outcomes in addition to posttreatment outcomes and compare the differences between the treated and control units (in our case, states with initiatives and states without them) before and after the treatment is administered. Any researcher attempting to use historical data to study behavior across states quickly becomes aware of the fact that even in the best-case scenario with a minimal number of data limitations, he or she will never be able to collect all the necessary data to account for important differences that exist among the states. Thus, there will almost always be unobserved effects on the dependent variable that need to be accounted for when observing variation in behavior across units of study.

The first half of our article compares three treatments representing competing theories of initiative effects on direct democracy. These three treatments allow us to elucidate when and how initiatives increase turnout: the effect of having adopted the initiative process but not yet using it, the effect of having used initiatives in the past but not using them in the current election, and the effect of having one or more initiatives on the ballot. Our control group in this case is the twenty-six states without the citizen initiative in their constitutions.

Since the treatments are not randomly assigned to the twenty-four states with the initiative process, we compare these effects using a differences-in-differences (DiD) design. Each treatment effect is calculated as follows: the average difference in turnout between the treatment and control states in elections before the treatment is administered is subtracted from the difference in turnout between the same two groups of states after the treatment is administered (for a simple explanation of DiD, see Meyer 1995). This calculation essentially washes away any differences between the two groups of states that vary slowly over time (or not at all). Instead of comparing the differences between the groups in the election just before the introduction and first use of initiatives and the election right after, we use multiple elections before and after each respective treatment is given. This design is much stronger against threats to internal and external validity than the basic DiD model (Meyer 1995).

In the second half of the article we begin to explore the differential effects of two different types of initiatives (competitive and uncompetitive initiatives) to more fully test the mobilization theory. We employ a DiD design here with multiple measures of our treatment effects: the number of competitive and the number of uncompetitive initiatives. The treatment effect equals the average effect each competitive (or uncompetitive) initiative has on turnout compared to states that do not have them on the ballot. A few factors bolster this design’s internal validity. Measuring the treatment repeatedly allows us to be certain that any observed effects of having competitive (or uncompetitive) initiatives on the ballot increase the validity of our inferences (Shadish, Cook, and Campbell 2002). Also, a different set of states have initiatives (both competitive and uncompetitive) on the ballot from election to election. This allows us to isolate the effects of such initiatives on turnout over time, further minimizing threats to internal validity. Finally, having such a long time series and a high number of repeated measurements guards against historical threats to validity (Shadish, Cook, and Campbell 2002).

We chose 1870 as the starting point of our analysis for two reasons; the first reason is a methodological one. To capture enough elections prior to most states’ adoption of the citizen initiative in the early twentieth century, we went back roughly ten elections to estimate a pretest measure of turnout. To best exploit a DiD design, it is important to have the “pretest” data be as similar to the “posttest” data as possible. The development of the Republican Party and the current partisan divide began in 1860 (Sundquist 1983). In addition, the ratification of the Fifteenth Amendment in 1870, allowing African Americans to vote, provides us a starting point for the current party system. The second reason for using 1870 as a starting point is a practical one. The data available prior to 1870 are inconsistent, especially for the dependent variable (VEP) and a key control variable (political competition).

Granted, neither component of our hybrid research design is a miracle cure, as analysts may not completely control for self-selection into the treatment group. To alleviate this concern, we use state fixed effects to control for the unobserved time invariant characteristics of each state. We also control for variation in some factors that have been shown to be influential in the adoption of the citizen initiative process and turnout: party competition (D. A. Smith and Fridkin 2008) and ethnic and geographic diversity (Bridges and Kousser, forthcoming). We control for autocorrelation and serial correlation using Driscoll–Kraay standard errors.¹

Operationalization of the Variables

The dependent variable for all of the models is the statewide turnout rate for the voting-eligible population (VEP) (McDonald and Popkin 2001). Voting-age population (VAP) as a measure of turnout may serve as a more theoretically tractable concept for studies of democratic theory; however, a consistent and accurate measure of the percentage of the eligible statewide population turnout enables a more precise estimation of the effect of direct democracy (Tolbert and Smith 2005). For VEP we use data from Rusk (2001) for 1870–1978 and data from McDonald (2009) for 1980–2008. Previous research (Tolbert and Smith 2005) has used McDonald’s VEP data; for consistency we use those same data when they are available.² We use turnout for the highest office; in presidential elections, it is obviously the presidential races. In midterm elections, we supplement Rusk’s congressional VEP data with Burnham, Clubb, and Flanigan’s (1991) senate and then gubernatorial election turnout if those races are present; otherwise, congressional VEP turnout is the measure.

We look at the introduction, usage, and competitiveness of initiatives during federal election years (presidential and midterm years) and test their effects on turnout. Through 2008, there have been 1,970 citizen initiatives voted on in federal general elections since 1904, when Oregonians first cast a ballot for altering local liquor laws and creating direct primaries. Of those, 455 were “competitive” races, slightly over 23 percent of all initiative races. The two most common measures of competitiveness in studies of congressional elections are absolute margins of victories of 10 percent and 20 percent (Jacobson 2004). For our purposes, we classify competitive races as having an absolute margin of victory of 10 percentage points or fewer.³ Data for initiative usage were provided by the National Conference of State Legislatures (2010) and the Initiative and Referendum Institute (2009) at the University of Southern California.

We measure the effects of mobilization using competitiveness as a proxy. For our purposes here, an initiative election is considered competitive if the winning side’s margin of victory was ten or fewer points. We acknowledge that this is an imperfect proxy to measure campaign intensity, but for a historical analysis, it is the best available measure. We also argue that expediency is not the only reason for using election results as a proxy; it is conceptually consistent with the logic of how campaigns allocate their resources (see the discussion of campaigns in the previous section).³

Specifically, we model mobilization effects using the square root of the number of competitive and uncompetitive initiatives because of the curvilinear relationship of the data. We agree with Tolbert, Bowen, and Donovan’s (2009) theoretical expectations of diminishing marginal effects as the number of initiatives on the ballot increases. They used simply the number of initiatives on the ballot, or the number of initiatives and the number of initiatives on the ballot squared (D. A. Smith and Tolbert 2004; Tolbert and Smith 2005; Tolbert, Grummel, and Smith 2001). We argue that these specifications do not match their theoretical claims of how the increasing number of initiatives affects turnout. Simply counting the number of initiatives leads to a linear prediction of turnout, something that is empirically questionable. Studies that have used the number of initiatives (with a positive coefficient) and the squared number of initiatives (with a negative coefficient) to predict turnout model a quadratic relationship. Those studies predict that as the number of initiatives increases beyond a certain point, turnout will decline. These predictions are models of voter fatigue. While there is evidence of ballot roll off and initiative-specific abstention when the number and complexity of ballot propositions increase (Bowler and Donovan 1998; Bowler, Donovan, and Happ 1992; Nichols 1998), electoral competition does not drive down aggregate rates of political participation. Since we argue that initiatives have a mobilizing impact with decreasing marginal returns, we believe using the square root of the number of initiatives (both competitive and uncompetitive) is the more theoretically tractable approach.

In addition to running our models with the square root of the number of competitive and uncompetitive initiatives, we ran our models using the number and number squared. None of the models (competitive or uncompetitive initiatives in presidential or midterm elections) had a statistically significant result for the number squared term. We believe that previous research has suffered from a truncation in the data for the number of initiatives on the ballot. When the coefficient of the number of initiatives is relatively large compared to the coefficient of the number of initiatives squared and the actual number of initiatives is small (over 99 percent of the cases), the models predict roughly the same positive curvilinear relationship. It is only when the number of initiatives gets large that the curvilinear relationship becomes quadratic and begins to predict large decreases in turnout (see Figure 1). We control for social and institutional context, specifically party competition, education, foreign-born and nonwhite populations, the secret ballot, women’s suffrage, and the presence of senatorial and gubernatorial elections. Please see the supplemental appendix (available at http://prq.sagepub.com/supplemental/) for a description of these variables.

Figure 1.

Predicted increase in turnout as a function of the number of initiatives on the ballot in midterm elections 1870–2008, using curvilinear model versus quadratic

Findings

The empirical analysis proceeds in two parts. First, in Table 1, we present regression results that demonstrate the existence of direct democracy is not enough to generate increases in turnout. Similarly, we show that the theory of a participatory spillover effect for increasing turnout is not supported by evidence. These two findings clearly indicate that the institution itself does not translate into more voters at the polls. In addition, the presence of initiatives on the ballot is not enough to bump up turnout in presidential elections. However, the initiatives on the ballot do increase turnout in midterm elections, further suggesting that campaign mobilization efforts are the driving force getting people into the voting booth. The second section of our findings, presented in Table 2, looks into this relationship more directly. By breaking down initiatives into competitive and uncompetitive categories, we present additional evidence that campaign mobilization is the root cause of turnout increases associated with direct democracy.

Table 1.

The Effect of Citizen Initiatives on Voter-Eligible Turnout, 1870–2008

Variable	Presidential	Midterm
Initiative process adopted, but not used	0.54 (2.03)	−0.34 (1.80)
Initiative process used in past, but not on the ballot	2.76 (1.68)	1.36 (1.77)
Initiatives on the ballot	2.25 (1.64)	3.76* (1.55)
Party competition	0.39*** (0.03)	0.37*** (0.04)
Education	8.53** (2.76)	2.11 (1.90)
Foreign-born citizens %	0.15 (0.11)	0.12 (0.11)
Nonwhite citizens %	−0.13 (0.15)	−0.22 (0.15)
Urban %	−0.03 (0.08)	−0.11 (0.07)
Secret ballot	−5.44* (2.08)	−7.18** (2.51)
Women’s suffrage	−7.58*** (1.96)	−7.85*** (1.73)
Gubernatorial election	1.49** (0.48)	−2.1 (1.14)
Senatorial election	0.27 (0.44)	0.22 (0.74)
Constant	36.94*** (5.04)	37.35*** (4.91)
Observations	1,629	1,612
Number of states	50	50
Within-R²	.49	.55

Driscoll–Kraay standard errors are in parentheses.

p < .05. **p < .01. ***p < .001.

Table 2.

The Effect of Competitive versus Uncompetitive Initiatives on Voter-Eligible Turnout, 1872–2008

	Presidential		Midterm
Variable	Competitive	Uncompetitive	Competitive	Uncompetitive
Square root of competitive initiatives	0.14 (0.43)		1.77*** (0.49)
Square root of uncompetitive initiatives		0.42 (0.52)		1.19* (0.54)
Party competition	0.39*** (0.03)	0.39*** (0.03)	0.37*** (0.04)	0.37*** (0.04)
Education	8.64** (2.67)	8.66** (2.67)	2.72 (1.94)	2.51 (1.88)
Foreign-born citizens %	0.14 (0.11)	0.14 (0.11)	0.1 (0.10)	0.1 (0.10)
Nonwhite citizens %	−0.13 (0.15)	−0.13 (0.15)	−0.21 (0.15)	−0.21 (0.15)
Urban %	−0.02 (0.07)	−0.02 (0.07)	−0.1 (0.06)	−0.11 (0.06)
Secret ballot	−5.65** (1.98)	−5.63** (2.00)	−7.35** (2.48)	−7.33** (2.48)
Women’s suffrage	−7.16*** (1.83)	−7.20*** (1.85)	−7.48*** (1.61)	−7.52*** (1.63)
Gubernatorial election	1.52** (0.51)	1.54** (0.52)	−2.12 (1.11)	−2.14 (1.12)
Senatorial election	0.37 (0.42)	0.34 (0.42)	0.32 (0.75)	0.35 (0.75)
Constant	37.03*** (4.95)	37.04*** (4.97)	37.06*** (4.91)	37.22*** (4.87)
Observations	1,629	1,629	1,612	1,612
Number of groups	50	50	50	50
Within-R²	.49	.49	.55	.55

Driscoll–Kraay standard errors are in parentheses.

p < .05. **p < .01. ***p < .001.

Before discussing the results in detail, it is useful to provide a few words on interpreting the key coefficients from the models. As explained in the previous sections, we employ a DiD design to test the effects of initiatives on turnout. Each model utilizes state fixed effects, which measure the deviation in turnout from the mean over time for that state as a function of our treatment variables in each time period (for a longer discussion of how to interpret fixed effects, see Beck 2001). Thus, the coefficients for our three key independent variables in Table 1 (having adopted the initiative process but not yet using it, having used the process in the past but not in the current election, having initiatives on the ballot) reflect an average of within-state averages. The models measure the average treatment effect of having initiatives (or not) on the ballot over time in each state, and the coefficients produced for the fifty-state regression model represent an average of the treated states’ average effects over time.

As Table 1 shows below, there is no support for either Hypothesis 1 or Hypothesis 2, the existence of the initiative process does not have any effect on voter turnout. States that have adopted the initiative process but do not have initiatives placed on the ballot fail to experience significantly higher turnout. This nonfinding is consistent across both presidential and midterm elections. We also cast doubt on claims that turnout in a given election is a function of using the initiative process in the past. States that have used the initiative in past elections but do not have an initiative on the ballot in a given election do not experience higher turnout in a given election, contradicting claims made by Hero and Tolbert (2004). The results in Table 1 do support Hypothesis 3 and the consistent finding in the literature that turnout in midterm elections is increased by the presence of initiatives on the ballot. On average, having initiatives on the ballot in midterm elections leads to an approximate increase of 3.7 percentage points in turnout in a state.

The regression results in Table 1 (midterm elections and presidential elections) provide solid evidence for our theory of mobilization. When there are no initiatives on the ballot, turnout is not affected, but when initiatives are present during favorable political conditions, they can increase turnout. Midterm elections are relatively low-information environments, and citizens are less engaged in politics than they are in presidential years (Jacobson 2004). Midterm elections provide an opportunity for statewide proposition campaigns to get their message out and mobilize the marginal voter. Presidential election years, on the other hand, are the most intense election environments, and the high intensity of those campaigns swamp the influence that ballot propositions have on turnout in those elections.

Control variables in both the midterm and presidential elections reinforce traditional findings in the literature about the role of institutional and political contexts’ effects on turnout. Women’s suffrage and the secret ballot significantly decreased voting eligible turnout. The expansion of suffrage to women sharply reduced turnout by more than seven percentage points. Similarly, the introduction of the secret ballot also abruptly decreased turnout by more than five percentage point in presidential elections and over seven percentage points in midterm elections. These two examples should serve as a stark warning that institutional changes can have unintended consequences on the political system. The other key control variable, party competition, has a strong positive effect on turnout. Measuring the change in turnout as party competition goes from a standard deviation below its mean to one standard deviation above it, turnout increases by sixteen points in presidential elections and by approximately fifteen points in midterm elections.⁴

We test for mobilization effects and find that the number of competitive initiative elections (margin of victory of ten or fewer percentage points) increases turnout and does so at a greater rate than the number of noncompetitive initiative elections. The regression results presented in Table 2 show that competitive and uncompetitive initiatives alike provide for a boost in turnout during in midterm elections, supporting Hypothesis 4. This is important because we include data from the introduction of the political institution, not just the past thirty years. However, as we saw in Table 1, the number of initiatives (regardless of the level of competition) is not significant in the presidential models, indicating that the most competitive ballot propositions are unable to mobilize voters beyond what the national campaigns can do.⁵

Figure 1 plots the predicted marginal effects for competitive initiatives and uncompetitive initiatives in midterm elections. It also includes a comparison of the quadratic model that uses the number of initiatives (with a positive coefficient) minus the number of initiatives squared (see Tolbert and Smith 2005). Competitive initiatives have the greatest effect on turnout, while uncompetitive initiatives increase turnout to a lesser degree.

Our results for midterm elections bolster the claim made by Tolbert, Bowen, and Donovan (2009) that it is campaign activity that drives turnout. Our other key finding is that this mobilization is not successful in presidential elections. Contrary to the analyses of Tolbert and Smith (2005), no form of ballot proposition provides a boost for turnout at traditional levels of statistical significance (p < .05). We argue that our historical analysis encapsulates a more accurate depiction of the effects of the role of direct democracy on turnout over the course of its existence. While we would love to measure campaign activity with expenditures, these data simply do not exist, and competitiveness is the closest proxy available.

The curvilinear effect presented in Figure 1 is modeled using the square root of the number of each type of proposition on the ballot (competitive and uncompetitive initiatives). Though the marginal effects show potentially large increases in turnout, up to 8 percentage points when there are 20 competitive initiatives, the maximum number of observed competitive initiative races is only 8, and that predicts a 5 percentage point jump in turnout. However, the mean number of competitive initiatives, given that any initiative is on the ballot, is less than one (0.71) competitive initiative per election, which predicts a 1.5 percentage point increase in turnout. There was a maximum of 21 uncompetitive initiatives in our sample, which predicts about a 5 percentage point increase in turnout at the maximum. Again, given that there are any initiatives on the ballot, the mean number of uncompetitive initiatives is only 2.31, which predicts a modest 1.8 percentage point increase in turnout.

Discussion and Conclusion

In this article, we have accomplished three things that elucidate whether and how direct democracy has had unintended consequences on voter turnout. First, we show that contrary to claims previously made in the literature, having direct democracy does not in and of itself lead to higher turnout, nor does having used the initiative process in past elections. We compare past versus current usage of the initiative and show that such claims are not true. Studies that have made claims about the initiative as an institution causing a permanently more engaged electorate should be looked at with a skeptical eye. Using the initiative process in the current election is what matters; states with initiatives on the ballot in midterms have higher turnout than states that do not (direct democracy does not increase turnout in presidential races). Second, we illustrate that mobilization is what drives turnout attributed to citizen initiatives with competitive ballot races affecting turnout more than noncompetitive races. Finally, we are the first to do test these theories using historical turnout data that cover a large enough time span to encapsulate the entire existence of the initiative process in the American states.

Our data and analyses do have limitations. We would love to extend Tolbert, Bowen, and Smith’s (2009) work using campaign expenditure data throughout our time series; of course this is not possible. Campaign finance data before the 1970s are not credible, and state-level campaign finance data are not easily available prior to the early 1990s. Therefore, we simply cannot construct a variable that directly measures campaign activity for every initiative prior to the 1990s. On the other hand, we are able to analyze turnout before and after a state adopts and begins using initiatives, an advantage that previous studies on initiatives and turnout do not enjoy.

Our findings show that the campaign environment of ballot propositions has direct effects on turnout. Though we believe we have answered several questions about the precise role that ballot propositions have on turnout, others remain. What types of initiatives are most likely to be competitive and hence drive up turnout? Do certain forms of campaign activity work better for direct democracy than candidate elections? Are referenda campaigns as effective in mobilizing the electorate as citizen initiatives? These questions are just a few of the many unsolved puzzles that remain regarding direct democracy’s effects on participation. Nonetheless, we do provide a reasonable amount of evidence that turnout is positively affected by the presence of competitive and uncompetitive (though less so) initiatives in midterm elections.

Footnotes

Acknowledgements

We owe many thanks to Gary Jacobson, Thad Kousser, Amy Bridges, Shaun Bowler, Lonna Atkeson, Jason McDaniel, Amanda Keller, Frederick Boehmke, Langche Zeng, Simon Jackman, Isaac Martin, John Ayers, and the participants of the Analysis of Aggregate Survey and Electoral Data Seminar at the University of California, San Diego for their invaluable feedback and encouragement throughout this project. We also thank the three anonymous reviewers who offered helpful and insightful commentary on previous drafts of this article. Finally, we thank the National Conference on State Legislatures and the Initiative and Referendum Institute for sharing their data on ballot propositions. All errors are of course our own.

The author(s) declared no potential conflicts of interests with respect to the authorship and/or publication of this article.

The author(s) received no financial support for the research and/or authorship of this article.

Notes

References

Abramowitz

Alan

. 2004. Terrorism, gay marriage, and incumbency: Explaining the republican victory in the 2004 election. Forum 2:Article 3. http://www.bepress.com/forum/vol2/iss4/art3.

Beck

Nathaniel

. 2001. Time-series–cross-section data: What have we learned in the past few years? Annual Review of Political Science 4:271–93.

Blair

Gus

. 1916. Letter from Gus Blair to Thomas J. Walsh. Thomas J. Walsh Papers, Library of Congress, Washington, DC.

Bowler

Shaun

Donovan

Todd

. 1998. Demanding choices: Opinion, voting, and direct democracy. Ann Arbor: University of Michigan Press.

Bowler

Shaun

Donovan

Todd

. 2002. Democracy, institutions, and attitudes about citizen influence on government. British Journal of Political Science 32:371–90.

Bowler

Shaun

Donovan

Todd

Happ

Trudi

. 1992. Ballot propositions and information costs: Direct democracy and the fatigued voter. Western Political Quarterly 45:559–68.

Bridges

Amy

Kousser

Thad

. Forthcoming. When do politicians give power to the people? Adoption of the citizen initiative in American states. State Politics & Policy Quarterly.

Burnham

W. Dean

Clubb

Jerome M.

Flanigan

William

. 1991. State-level congressional, gubernatorial, and senatorial election data for the United States, 1824–1972 [Computer file] (Study No. 75). Ann Arbor, MI: Inter-University Consortium for Political and Social Research. doi:10.3886/ICPSR0007510.3886/ICPSR00075

Childers

Matthew

Popkin

Samuel L.

2007. Has media circus ruined our bread? Paper presented at the annual meeting of the American Political Science Association, Chicago.

10.

Downs

Anthony

. 1957. An economic theory of democracy. New York: Harper & Row.

11.

Dyck

Joshua J.

Seabrook

Nicholas R.

2010. Mobilized by direct democracy: Short-term versus long-term effects and the geography of turnout in ballot measure elections. Social Science Quarterly 91:188–208.

12.

Ernst

Howard R.

2001. The historical role of narrow-material interests in initiative politics. In Dangerous democracy: The battle of ballot initiatives in America, ed. Sabato

Larry J.

Larson

Bruce A.

, 1–25. Oxford, UK: Rowan & Littlefield.

13.

Fowler

James H.

Baker

Laura A.

Dawes

Christopher T.

2008. Genetic variation in political participation. American Political Science Review 102:233–48.

14.

Fowler

James H.

Dawes

Christopher T.

2008. Two genes predict voter turnout. Journal of Politics 70:579–94.

15.

Gelman

Andrew

Hill

Jennifer

. 2006. Data analysis using regression and multilevel/hierarchical models. New York: Cambridge University Press.

16.

Gerber

Alan S.

Green

Donald P.

Shachar

Ron

. 2003. Voting may be habit-forming: Evidence from a randomized field experiment. American Journal of Political Science 47:540–50.

17.

Goldstein

Ken

Freedman

Paul

. 2002. Campaign advertising and voter turnout: New evidence for a stimulation effect. Journal of Politics 64:721–40.

18.

Heckelman

Jac C.

1995. The effect of the secret ballot on voter turnout rates. Public Choice 82:107–24.

19.

Hero

Rodney E.

Tolbert

Caroline J.

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.

20.

Initiative and Referendum Institute. 2009. Statewide initiatives 1904–2008. http://www.iandrinstitute.org/Usage.htm (accessed May 1, 2009).

21.

Jacobson

Gary C.

2004. The politics of congressional elections. 6th ed. New York: Pearson/Longman.

22.

Karp

Jeffrey A.

Banducci

Susan A.

2000. Going postal: How all-mail elections influence turnout. Political Behavior 22:223–39.

23.

Key

V. O.

Jr.

1949/1984. Southern politics in state and nation. Knoxville: University of Tennessee Press.

24.

Kousser

J. Morgan

. 1974. The shaping of southern politics: Suffrage restriction and the establishment of the one-party south, 1880–1910. New Haven, CT: Yale University Press.

25.

Kousser

Thad

Mullin

Megan

. 2007. Does voting by mail increase participation? Using matching to analyze a natural experiment. Political Analysis 15:428–45.

26.

Lupia

Arthur

. 1994. Shortcuts versus encyclopedias: Information and voting behavior in California insurance reform elections. American Political Science Review 88:63–76.

27.

Magleby

David B.

1984. Direct legislation: Voting on ballot propositions in the United States. Baltimore: Johns Hopkins University Press.

28.

Marcus

Robert D.

1971. Grand old party: Political structure in the gilded age 1880–1896. New York: Oxford University Press.

29.

McDonald

Michael P.

2009. Turnout 1980–2008. http://elections.gmu.edu/Turnout%201980-2008.xls (accessed April 1, 2009).

30.

McDonald

Michael P.

Popkin

Samuel L.

2001. The myth of the vanishing voter. American Political Science Review 95:963–74.

31.

Meyer

Bruce D.

1995. Natural and quasi-experiments in economics. Journal of Business & Economic Statistics 13:151–61.

32.

National Conference of State Legislatures. 2010. Ballot measure database. http://www.ncsl.org/default.aspx?tabid=16580 (accessed January 27, 2010).

33.

Nichols

Stephen M.

1998. State referendum voting, ballot roll-off, and the effect of new electoral technology. State and Local Government Review 30:106–17.

34.

Nicholson

Stephen P.

2003. The political environment and ballot proposition awareness. American Journal of Political Science 47:403–10.

35.

Nicholson

Stephen P.

2005. Voting the agenda: Candidates, elections, and ballot propositions. Princeton, NJ: Princeton University Press.

36.

Patterson

Thomas E.

2002. The vanishing voter: Public involvement in an age of uncertainty. New York: Knopf.

37.

Popkin

Samuel L.

1994. The reasoning voter: Communication and persuasion in presidential campaigns. 2nd ed. Chicago: University of Chicago Press.

38.

Putnam

Robert D.

1995. Tuning in, tuning out: The strange disappearance of social capital in America. PS: Political Science and Politics 28:664–83.

39.

Rusk

Jerrold G.

2001. A statistical history of the American electorate. Washington, DC: CQ Press.

40.

Schlozman

Daniel

Yohai

Ian

. 2008. How initiatives don’t always make citizens: Ballot initiatives in the American states, 1978–2004. Political Behavior 30:469–89.

41.

Shadish

William R.

Cook

Thomas D.

Campbell

Donald T.

2002. Experimental and quasi-experimental designs for generalized causal inference. Boston: Houghton Mifflin.

42.

Shaw

Daron R.

2006. The race to 270: The electoral college and the campaign strategies of 2000 and 2004. Chicago: University of Chicago Press.

43.

Smith

Daniel A.

DeSantis

Matthew

Kassel

Jason

. 2006. Same-sex marriage ballot measures and the 2004 presidential election. State and Local Government Review 38:78–91.

44.

Smith

Daniel A.

Fridkin

Dustin

. 2008. Delegating direct democracy: Interparty legislative competition and the adoption of the initiative in the American states. American Political Science Review 102:333–50.

45.

Smith

Daniel A.

Tolbert

Caroline J.

2004. Educated by initiative: The effects of direct democracy on citizens and political organizations in the American states. Ann Arbor: University of Michigan Press.

46.

Smith

Mark A.

2001. The contingent effects of ballot initiatives and candidate races on turnout. American Journal of Political Science 45:700–706.

47.

Sundquist

James L.

1983. The dynamics of the party system. Washington, DC: Brookings Institution.

48.

Tolbert

Caroline J.

Bowen

Daniel C.

Donovan

Todd

. 2009. Initiative campaigns: Direct democracy and voter mobilization. American Politics Research 37:155–92.

49.

Tolbert

Caroline J.

Grummel

John A.

Smith

Daniel A.

2001. The effects of ballot initiatives on voter turnout in the American states. American Politics Research 29: 625–48.

50.

Tolbert

Caroline J.

Smith

Daniel A.

2005. The educative effects of ballot initiatives on voter turnout. American Politics Research 33:283–309.