DIRECT DEMOCRACY, PUBLIC OPINION, AND CANDIDATE CHOICE

Transcription

1 Public Opinion Quarterly, Vol. 74, No. 1, Spring 2010, pp DIRECT DEMOCRACY, PUBLIC OPINION, AND CANDIDATE CHOICE DANIEL A. SMITH* CAROLINE J. TOLBERT Abstract We argue that the rich information environment created by ballot measures makes some policy issues more salient, shaping voters positions on broad topics such as the importance of the economy. This in turn may affect candidate choice for national and statewide elected office. We theorize that the creation of state-specific issue publics may be the causal mechanism underlying this process. Using large-sample national survey data with robust samples from the 50 U.S. states, we test whether mass support for a specific policy raising the minimum wage is higher in states where the issue is on the ballot, whether being directly exposed to initiative campaigns elevates the importance of broad issues like the economy, and whether the economic-related ballot measures prime support for Democratic candidates. We find that exposure to minimum-wage ballot measure campaigns in 2006 modified support for the policy among partisan subsamples (with Democrats becoming more likely and Republicans less likely to support the measure), increased the saliency of the economy in general among these targeted populations, and primed support for Democratic candidates up and down the ballot. Direct Democracy, Public Opinion, and Candidate Choice If statewide ballot measure campaigns can prime candidate vote choice, might the initiative process concurrently engender stronger public preferences for certain issues? Recent studies have examined how statewide ballot measures can have an educative effect, increasing democratic competence, political DANIEL A. SMITH is with the Department of Political Science, University of Florida, Gainesville, FL, USA. CAROLINE J. TOLBERT is with the Department of Political Science, University of Iowa, Iowa City, IA, USA. The authors would like to thank Sarah Morehouse, Justin Phillips, Fred Boehmke, Ted Lascher, Josh Dyck, and the other participants at the 2008 American Political Science Association annual meeting for their helpful feedback on a previous version of this paper, and to Steve Nicholson for his unstinting encouragement.*address correspondence to Daniel A. Smith, Department of Political Science, PO Box , Anderson Hall, University of Florida, Gainesville, FL 32611, USA; dasmith@ufl.edu. doi: /poq/nfp097 Advance Access publication February 16, 2010 The Author Published by Oxford University Press on behalf of the American Association for Public Opinion Research. All rights reserved. For permissions, please journals.permissions@oxfordjournals.org

2 86 Smith and Tolbert interest, and participation (Smith and Tolbert 2004), as well as help set policy agendas and prime vote choice for candidates (Nicholson 2005; Donovan, Tolbert, and Smith 2008; Campbell and Monson 2008; Smith, DeSantis, and Kassel 2006; Bowler, Segura, and Nicholson 2006). Scholars, though, have yet to fully examine how exposure to ballot measures may have context-specific campaign effects that may alter public opinion in the short term (Nicholson 2008). Drawing on the initial insights of Converse s (1964) theory of issue publics and the importance of elite discourse in the contingency of public opinion (Zaller 1992), we argue that exposure to direct democracy campaigns may affect the way individuals (partisans and otherwise) think about issues, leading to heterogeneity (and volatility) of opinion formation across the states (see, e.g., Bowler and Donovan 1998). In addition to setting the agenda, we suggest that the rich information environment created by ballot measures may also make some issues more salient, shaping voters positions on broad topics such as the importance of the economy. This in turn may prime candidate choice for national and statewide elected office. Here, we examine the impact of initiatives placed on statewide ballots in the 2006 midterm elections aimed at raising the minimum wage. Spurred on by Congress s unwillingness to raise the federal minimum wage above $5.15 an hour (in place since 1997), progressive organizations decided to take matters into their own hands. Led by a loose coalition of organized labor, the Ballot Initiative Strategy Center (BISC), and the Association of Community Organizations for Reform Now (ACORN), the coalition placed initiatives raising the minimum wage on the ballots of six states (AZ, CO, MO, MT, NV, and OH). All six measures passed, with an average passage rate of 65 percent. Substantively, the six ballot initiatives were quite similar, as were the dynamics of the campaigns. 1 For many liberals, the minimum wage initiatives represented an economic counterpunch to the conservative same-sex marriage 1. In Arizona, proposition 202 raised the minimum wage from $5.15 an hour to $6.75 an hour, with cost-of-living adjustments. The measure passed with 65.4 percent of the vote. Money raised for and against the measure was fairly even, with proponents taking in $1.4 million and opponents raising $1.1 million. Amendment 42 in Colorado bumped the state s minimum wage to $6.85 per hour, with future annual adjustments tied to the Consumer Price Index. The measure received 53 percent of the vote. Proponents raised $1.2 million, but were outspent 2 to 1 by opponents. Proposition B in Missouri passed with 76 percent of the vote, increasing the state s minimum wage from $5.15 an hour to $6.50 an hour, with similar cost-of-living adjustments. Proponents raised $1.8 million, far surpassing the $149,900 raised by opponents of the measure. In Montana, initiative I-151 passed with 73 percent of the vote, raising the state s minimum wage to $6.15 an hour with annual cost-of-living adjustments. Opponents were outspent 3 to 1, with proponents spending nearly $300,000. Nevada s question 6, which passed with 69 percent of the vote, increased the minimum wage to $6.15 an hour for employers not providing health benefits. Opponents raised more than $350,000, three times more than the proponents. Finally, in Ohio, issue 2 passed with 56 percent of the vote, elevating the minimum wage to $6.85 an hour and indexing it to inflation. Proponents raised $3.6 million, twice as much as the opponents.

3 Direct Democracy and Public Opinion 87 measures that possibly tipped the 2004 presidential election to President George W. Bush. 2 Taking advantage of the very large sample of the 2006 Cooperative Congressional Election Study (CCES) survey, and merging the state-level variable of whether an initiative for raising the minimum wage was on a state s ballot, we examine mass opinion on the issue of raising the minimum wage and on the economy more generally. We begin by testing whether Democrats, Republicans, and Independents directly exposed to statewide campaigns pushing for an increase in the minimum wage had different levels of support for the public policy compared to similar partisans living in states without a measure on the ballot. We then examine whether being exposed to minimumwage ballot measures affected the weight (or salience) respondents assigned to agreeing that the economy was the most important issue in the election. Specifically, we are interested in whether the minimum-wage measures primed Democrats to think about economic issues, thereby elevating the economy as the most important problem facing the country. Finally, building on the work of Nicholson (2005) and Donovan, Tolbert, and Smith (2008), we are interested in whether direct exposure to ballot measure campaigns shapes candidate vote choice. We test whether individuals who favored raising the minimum wage and who resided in a state with a corresponding ballot measure were more likely to vote for the Democratic candidates running for U.S. House of Representatives, U.S. Senate, and governor. HOW DO BALLOT MEASURES AFFECT ELECTIONS? CREATING STATE-SPECIFIC ISSUE PUBLICS Recent research by Nicholson (2005) and others reveals how ballot measures can have spillover effects that can alter the electoral agenda and prime voters to support (or oppose) both federal and state candidates. Nicholson contends that because the two major parties own certain issues and voters rely on partisan stereotypes, priming of candidate choice can occur regardless of whether a candidate has taken a stance on the ballot measure. Voters may not necessarily link issues on the ballot directly to specific candidates; yet, ballot propositions may have indirect spillover effects that indiscriminately affect citizens political judgments in candidate elections, regardless of whether candidates have spoken in favor of or against measures on the ballot (Nicholson 2005, p. 128). Donovan, Tolbert, and Smith (2008) find that in the 2004 presidential contest, ballot measures banning same-sex marriage may have had priming effects on the presidential contest benefiting conservatives. What scholars have yet to nail down, though, is why and how some ballot 2. According to Kristina Wilfore, former Executive Director of BISC, the initiatives marked a moment when progressives finally realize[d] that ballot measures are an important political opportunity (Frank 2006).

4 88 Smith and Tolbert measures are able to prime certain voters to support (or oppose) candidates running for office, and whether such spillover effects extend beyond the moral values issue of gay marriage in the 2004 presidential election. Although we investigate here the spillover effects on candidate races from the 2006 minimum-wage ballot measures, we are interested primarily in unpacking how ballot measures on certain public policies may be able to shape public opinion in the short term. We suggest that some ballot measures are able to prime candidate vote choice among the mass electorate because they elevate the importance of broad issues like the economy or moral values. Because partisan identification shapes policy orientations (Campbell et al. 1960, pp ; Markus 1982) and enables issue stereotyping (Conover and Feldman 1989; Nicholson 2005), ballot measures may elevate the importance of certain issues for various segments of the mass electorate. If our supposition is correct, it is possible that the creation of state-specific issue publics may be the causal mechanism underlying how ballot measures are able to prime voters evaluation of candidates for elective office. Rather than assuming that the spillover effects of ballot measures on candidate races are the automatic byproducts of partisan stereotyping or generic campaign effects, we suspect that voters in 2006 who were directly exposed to ballot campaigns regarding an increase in the minimum wage were inclined to grant more weight to the importance of the economy (relative to other issues), when compared to voters not directly exposed to the ballot measures. WHY BALLOT MEASURES MAY SHAPE PUBLIC OPINION In his seminal piece, Converse (1964) finds that most Americans lack political sophistication and are thereby unable to maintain a meaningful and consistent set of attitudes. Unlike elites and politically interested citizens who possess higher political knowledge and play an important role in the creation of mass preferences (McClosky 1964), large numbers of individuals do not have coherent positions on issues; rather, they have non-attitudes, or meaningless answers they make up on the spot. Any ideological belief system most individuals might have, Converse contends, is fragmented into narrower, overlapping issue publics (see also Nie, Verba, and Petrocik 1976). Building on Converse s work, and motivated by Key s (1966, p. 8) observation that the electorate, however unsophisticated, is moved by concern about central and relevant questions of public policy, Krosnick (1990, p. 75) shows how the mass public is stratified in a patchwork quilt fashion, comprised of individuals who care deeply about a particular issue. Everybody is sophisticated, Dalton (2006, p. 27) quips, paraphrasing Will Rogers, only on different subjects. The scholarly jury, though, is still out as to whether individuals lack sophisticated and consistent attitudes on issues, or whether apparent issue instability is just a matter of measurement error (Converse 2006). Expanding beyond

5 Direct Democracy and Public Opinion 89 Converse s work, Hillygus and Shields (2008, p. 6) find that just because voters do not have constrained ideologies does not necessarily mean that they do not hold policy preferences. Recent empirical evidence challenges the notion that voters lack relatively stable policy preferences, as issue preferences inform candidate vote choice (Ansolabehere, Rodden, and Snyder 2008). Setting aside this larger debate regarding issue coherence and stability of belief systems in the mass public, we suggest that it is possible for ballot measures to help create cohesive (albeit ephemeral) state-specific issue publics. Why might direct exposure to ballot issue campaigns help temporarily solidify attitudes on particular issues, thereby consolidating issue publics within the states? In addition to boosting voter turnout and increasing political interest and knowledge (Smith and Tolbert 2004; Tolbert, Grummel, and Smith 2001; Tolbert, Bowen, and Donovan 2009), as well as framing issues and shaping vote preferences in candidate races (Nicholson 2005), it is possible that statewide ballot measure campaigns may also help shape statelevel public opinion. By increasing the salience of some issues, while ignoring others, Nicholson (2008, p. 67) reasons, direct legislation has the capacity to focus attention on an issue and make it politically relevant to the citizens of a state. Just as scholars have found that specific campaign appeals in candidate races may mobilize various issue publics when the topic is important to the individual (Converse 1964; Hillygus and Shields 2008; Hutchings 2003; Krosnick 1990), ballot measures may have the same effect. For example, micro-targeting cross-pressured partisans with policy appeals (Hillygus and Shields 2008) may be more effective if there is a ballot measure on the topic in the respondent s state. Direct democracy may help to lubricate and make more effective micro-targeting campaigns. Furthermore, from a psychological perspective, issues appear to become important to individuals if they perceive a policy to directly affect their material self-interest (Converse 1964), if they identify with a group that takes a strong and consistent position on the issue (Nelson and Kinder 1996), or if a policy is relevant to their basic social and personal values (Krosnick 1990, pp ). Irrespective of the psychological cause driving issue importance, it is possible that statewide ballot measures may shape public opinion on broader issues. We argue that being directly exposed to ballot measures may temporarily elevate an individual s engagement with a particular issue. The greater a person s level of cognitive engagement with an issue, the more likely he or she is to be exposed to and comprehend in a word, to receive political messages concerning that issue Zaller (1992, p. 42) argues, and the more recently a consideration has been called to mind or thought about, the less time it takes to retrieve that consideration from memory and bring it to the top of the head for use (p. 48). Furthermore, ballot measures because they force voters to take a definitive yes or no stance on issues may accentuate or reinforce an individual s set of predispositions or core values that is relatively stable and coherent (Feldman 1988; McClosky and Zaller

6 90 Smith and Tolbert 1984; Sniderman and Piazza 1993). As Converse (1964, p. 246) conceded, different controversies excite different people to the point of real opinion formation. Thus, if voters regularly use heuristics informational shortcuts to inform candidate preferences (Bowler and Donovan 1998; Lau and Redlawsk 2006; Lupia 1994; Popkin 1991) and if they are susceptible to elite cues when making these decisions (Zaller 1992), statewide ballot measure campaigns may have a hand in creating temporary state-level issue publics. It is important not to oversell the impact that targeted campaigning may have on augmenting the importance individuals place on a broad issue like the economy (see Karch and Sides 2008). 3 Most statewide ballot measures, however, tend to be much more extensive than targeted advertising in candidate campaigns. Extensive mass media and grassroots campaigns are common features of ballot measures. Initiative proponents and opponents utilize campaign consultants, signature-gathering firms, and paid advertising, and often involve legal controversies (Bowler, Donovan, and Tolbert 1998). Many initiative campaigns exceed the amount spent on gubernatorial campaigns; over $524 million was spent on 73 initiatives on the ballot in November In most states, registered voters receive pamphlets containing information about ballot issues, and some ballot proposition campaigns may even draw more media attention and spending than prominent candidate races (Magleby 1984). Bolstering our suspicion that ballot measures may indeed shape mass opinion, there is evidence that ballot measures targeting minorities (e.g., immigrants, gays and lesbians) can worsen public opinion about that group in the state (Wenzel, Donovan, and Bowler 1998), but few have followed up on this early research as we do here. We submit that the 2006 statewide ballot initiatives increasing the minimum wage spurred opinion formation among segments of the mass public, thereby priming candidate vote support. Hypotheses, Data, and Methods In examining the effects of direct democracy state-specific public opinion, the saliency of opinions, and candidate vote choice, we test three interrelated hypotheses: 1) Exposure to state minimum-wage ballot measure campaigns increases support for the public policy among targeted partisan subsamples; 2) Exposure to minimum-wage ballot measures increases the saliency of the economy among partisan subsamples; 3) Exposure to economic minimumwage ballot measures primes Democratic candidate vote choice. We expand on previous research using American National Election Study (ANES) and Pew surveys by drawing on national opinion data from large and 3. For example, Karch and Sides (2008, p. 469) find little or no effect of issue-specific advertising in the 1998, 2000, and 2002 general elections targeting the turnout of three issue publics parents, veterans, and seniors. Their weak findings suggest that voters may not be more susceptible to campaign messages when those messages focus on the issues personally important to them (Karch and Sides 2008, p. 467).

7 Direct Democracy and Public Opinion 91 randomly drawn samples with all 50 states. Our data come from the 2006 CCES conducted by Polimetrix. 4 The survey sampled more than 36,500 respondents following the 2006 midterm elections with a common battery of questions comparable to the ANES or General Social Survey (GSS), but with sample sizes 15 times larger than these other omnibus surveys. This allows for relatively large samples within states (3,637 respondents from California alone), sufficient for assessing state contextual effects. Unlike the large-sample Current Population Surveys (CPS), the CCES includes a measure of partisanship, critical in testing our priming hypothesis, and also many opinion and vote choice variables not found in census data. The models are estimated by clustering respondents by the 50 states to adjust the standard errors for the multilevel data (Primo, Jacobsmeier, and Milyo 2007). 5 Testing our models using these survey data provides a more robust test of the effects of direct democracy on public opinion and vote choice than does previous research. Our empirical analysis proceeds in three steps, mirroring our hypotheses. In tables 1 and 2, we measure factors predicting opinions on a specific public policy. Our primary dependent variable is a binary response variable that measures support for increasing the federal minimum wage in the 2006 elections. Respondents who favor increasing the minimum wage are coded 1, with those not favoring coded 0. 6 In table 3, we expand the analysis by testing the effects of ballot measure exposure on the saliency of policy opinions. The dependent variable is opinion on the most important problem facing the nation. 7 The most common response not surprisingly, given that the 2006 election was considered by many to be a referendum on the Iraq War and President Bush s management of it was the war in Iraq, with 23.6 percent 4. This sample is constructed using a technique called sample matching. The researchers create a list of all U.S. consumers to generate a set of demographic characteristics that should be mirrored in the survey sample. Then, using a matching algorithm, the researchers select respondents who most closely resemble the consumer data from a pool of opt-in participants. The sample is stratified to ensure large samples within states. More information regarding sample matching is available at: These data were collected over a three-month period from September to November The models are estimated using Polimetrix survey weights. The response rate for the 2006 CCES was 93.1 percent (Vavreck and Rivers 2008). 5. Failing to cluster respondents by geographic area artificially reduces the size of the standard errors of the contextual variables (exposure to a minimum-wage ballot measure), which increases the chance of finding a statistically significant effect on the outcome variable when none exists (Primo, Jacobsmeier, and Milyo 2007). 6. In the 2006 CCES, this question was variable 2072; of the respondents, 26,523 favored increasing the minimum wage and 8,916 were opposed. Question wording: As you may know, the federal minimum wage is currently $5.15 an hour. Do you favor or oppose raising the minimum wage to $7.25 an hour over the next two years, or not? Responses included Oppose and Support. 7. In the 2006 CCES, this question was variable 2002; 2,216 respondents out of 36,128 mentioned the economy and jobs as the most important problem facing the nation. Question wording: What is the Most Important Problem facing the country today?

8 92 Smith and Tolbert Table 1. Factors Influencing Support for an Increase in the Minimum Wage by Partisan Subgroups Democrats Republicans Independents b p b p b p Independent Variables (RSE) (RSE) (RSE) Minimum-wage ballot measure state (.250) (.073) (.061) State median income (.000) (.000) (.000) Percent change gross state product, (2.12) (.585) (.541) Union member past or present (.166) (.067) (.087) Political interest (.142) (.070) (.067) Income (.027) (.012) (.017) Age (.006) (.003) (.003) Married (.112) (.071) (.072) Education (.058) (.023) (.029) Asian (.462) (.249) (.379) Latino (.169) (.093) (.089) Black (.257) (.196) (.119) Male (.106) (.065) (.081) Gay marriage most important issue (.352) (.161) (.113) Iraq most important issue (.202) (.098) (.116) Terrorism most important issue (.222) (.061) (.131) Economy most important issue (.169) (.209) (.188) Health care most important issue Retrospective evaluation of national economy (.234) (.137) (.179) (.116) (.041) (.059) Continued

9 Direct Democracy and Public Opinion 93 Table 1. Continued Democrats Republicans Independents b p b p b p Independent Variables (RSE) (RSE) (RSE) Retrospective evaluation of state economy (.154) (.036) (.072) Constant (.789) (.382) (.523) Pseudo R Log likelihood N 6,719 6,149 7,015 of respondents naming this issue. The second most frequently cited problem was terrorism, with 17.5 percent of the sample mentioning this concern. Corruption in government, health care and health costs, and immigration followed in frequency. Sixth among the issues mentioned, but nearly tied with immigration, was the economy and jobs. Although the frequency of this concern was relatively low in 2006, compared with 2008 when the economy dominated, we are interested in whether residing in a state with minimumwage measures on the ballot increased concerns about the economy. 8 Table 4 expands the analysis further, by testing whether respondents residing in states with minimum wage on the ballot were primed to vote for Democratic congressional and gubernatorial candidates, above and beyond concerns with the Iraq War. The dependent variables measure vote intentions in the 2006 election. 9 Three binary dependent variables are coded 1 for a Democratic candidate choice for U.S. House, U.S. Senate, and governor, and 0 for a Republican or other party vote choice. The primary explanatory variable measures exposure to a minimum-wage ballot measure campaign in a respondent s state. The six states with a 8. To check the robustness of our findings, we ran several alternative specifications where the sample consisted of only respondents who named one of the top six concerns, each with over 2,000 respondents: war in Iraq = 8,535, terrorism = 6,334, corruption in government = 4,081, immigration = 4,023, health care = 2,930, and economy and jobs = 2,216. After this, the most frequently cited response was other, with 1,937 responses. The same positive coefficient for residing in a state with a ballot measure was found with the reduced sample, when those saying the economy and jobs made up eight percent of the sample instead of six percent. Alternative specification, with other issues dropped to make the economy a larger percentage of the sample, revealed similar findings. A question on the second most important issue facing the nation was not asked. 9. In the 2006 CCES, variable 3054 measures whether the respondent intended to vote for the Democratic U.S. House candidate, variable 3056 a Democratic U.S. Senate candidate, and variable 3058 the Democratic candidate for governor. Question wording: For which candidate do you intend to vote in the race for U.S. House? Similar wording was used for the other two races.

10 94 Smith and Tolbert Table 2. Factors Influencing Support for an Increase in the Minimum Wage, Full Sample with Interaction Term Independent Variables b (RSE) Interaction model Democrat X minimum-wage ballot measure state (.233) Minimum-wage ballot measure state (.049) Democrat (.079) Republican (.059) State median income (.000) Percent change gross state product, (.497) Union member past or present (.051) Political interest (.048) Income (.009) Age (.002) Married (.044) Education (.018) Asian (.224) Latino (.071) Black (.117) Male (.047) Gay marriage most important issue (.091) Iraq most important issue (.069) Terrorism most important issue (.058) p Continued

11 Direct Democracy and Public Opinion 95 Table 2. Continued Interaction model b p Independent Variables (RSE) Economy most important issue (.107) Health care most important issue (.088) Retrospective evaluation of national economy (.030) Retrospective evaluation of state economy (.044) Constant (.314) Pseudo R 2.31 Log Likelihood N 19,883 minimum-wage initiative on the ballot in the 2006 election are coded 1, and respondents from the other 44 states are coded 0. In all, 4,471 respondents (12.2 percent) in our sample resided in states with a minimum-wage measure on the ballot. The second key explanatory variable is the respondent s partisanship. We expect Democrats to be more likely to think the economy is the most important problem if they reside in a state where the minimum-wage campaigns and the media focus attention on issues of jobs and substandard wages. Following previous research suggesting that independents have unique preferences on policy issues compared to partisans (Donovan, Parry, and Bowler 2005), we measure partisanship with a standard three-point measure. 10 This provides a conservative test, as Independents leaning Democrat are grouped with Independents, not with Democratic partisans. Because we are measuring opinions on economic policy and candidate voting, the models include controls for factors known to shape economic evaluations, including whether the respondent is a current or past union member, 11 average state median income using data from the 2000 U.S. Census Bureau, percent change in state gross product from 2000 to 2006, retrospective 10. In the 2006 CCES, variable 3005 is used, with 36,346 responses to this question. Question wording: Generally speaking, do you think of yourself as a...?, Democrat, Republican, Independent, other. In the sample, 32 percent of respondents identified as Democrats, 31 percent as Republicans, and 31 percent as Independents, with six percent reporting other. 11. In the 2006 CCES, variable 2082 measures whether the respondent is a current (8.5 percent) or past (26 percent) union member. These response options were combined to create a binary variable where 1 indicates union member and 0 otherwise.

12 96 Smith and Tolbert Table 3. Factors Influencing Respondent s Most Important Issue in the Election Economy most important issue Iraq most important issue Terrorism most important issue Health care most important issue b p b p b p b p Independent Variables (RSE) (RSE) (RSE) (RSE) Democrat X minimum-wage ballot measure state (.098) (.108) (.340) (.164) Minimum-wage ballot measure state (.191) (.074) (.077) (.148) State median income (.000) (.000) (.000) (.000) Democrat (.068) (.044) (.077) (.065) Republican (.082) (.047) (.041) (.066) Union member past or present (.084) (.040) (.043) (.036) Income (.015) (.007) (.006) (.012) Age (.002) (.001) (.002) (.001) Married (.046) (.035) (.042) (.049) Education (.032) (.014) (.014) (.024) Asian (.239) (.140) (.436) (.288) Latino (.102) (.082) (.068) (.080) Black (.106) (.062) (.112) (.096) Male (.060) (.037) (.052) (.054) Retrospective evaluation of national economy (.046) (.028) (.024) (.022) Retrospective evaluation of state economy (.078) (.028) (.025) (.027) Continued

13 Direct Democracy and Public Opinion 97 Table 3. Continued Economy most important issue Iraq most important issue Terrorism most important issue Health care most important issue b p b p b p b p Independent Variables (RSE) (RSE) (RSE) (RSE) Constant (.373) (.145) (.235) (.274) Pseudo R Log likelihood N 28,877 28,877 28,877 28,877 evaluations of the national economy, and retrospective evaluations of the state economy (Lewis-Beck 1988, 2006; Fiorina 1981). 12 We expect respondents with poor evaluations of the national and state economy to be more likely to support increasing the minimum wage, regardless of whether they live in a state with a ballot measure campaign. Tables 1 and 2 also control for the policy issue that the respondent thinks is the most important facing the nation in the election, which we then use as a dependent variable in table We expect those generally more interested in politics to have opinions on policy issues, such as increasing the minimum wage. 14 The models also include standard demographic controls for gender (males = 1, females = 0), age (in years), education (a six-point scale from no high school to postgraduate 12. Respondents were asked to evaluate the change in the nation s economy (and to evaluate the change in their state s economy) over the past year on a five-point scale from gotten much better (coded 1) to gotten much worse (coded 5). In the 2006 CCES, variable 3008 measures change in the nation s economy over the past year and variable 3009 measures change in the state economy over the past year. Question wording: Would you say that over the past year the nation s economy has.gotten much better, gotten better, stayed about the same, gotten worse, gotten much worse? In terms of national economic evaluations, 16 percent felt the economy was much better and 21 percent said gotten better, while 28 percent said the national economy had gotten worse and 14 percent said gotten much worse. 13. In the 2006 CCES, variable 2002 measures the most important issue facing the nation. A series of binary variables measure the top reasons given, including the Iraq War, terrorism, the economy, health care, and gay marriage. The latter is included, given that gay marriage ballot measures also occurred on a number of state ballots in 2006 and The survey instrument asked, How interested are you in politics and current affairs? The distribution of responses to this variable was such that we collapsed the responses so that 1 was coded very much interested and 0 was coded if the respondent said they were somewhat or not interested in politics.

14 98 Smith and Tolbert Table 4. Factors Influencing Democratic Vote Choice Democratic House Vote Democratic House Vote Democratic Senate Vote Democratic Governor Vote b p b p b p b p Independent Variables (RSE) (RSE) (RSE) (RSE) Favors increasing minimum wage X minimum-wage (.184) (.231) (.137) ballot measure state Minimum wage ballot measure state (.226) (.292) (.243) Favors increasing minimum wage (.075) (.081) (.081) (.109) Democrat (.049) (.049) (.079) (.071) Republican (.068) (.068) (.105) (.116) State median income (.000) (.000) (.000) (.000) Percent change gross state (.762) (.716) (1.516) (1.211) product Union member past or present (.046) (.046) (.066) (.054) Political interest (.039) (.039) (.082) (.078) Liberal (.049) (.049) (.071) (.096) Conservative (.075) (.075) (.072) (.107) Income (.010) (.010) (.012) (.013) Age (.002) (.002) (.002) (.002) Married (.044) (.045) (.061) (.068) Education (.014) (.014) (.021) (.024) Asian (.208) (.209) (.257) (.254) Latino Continued

15 Direct Democracy and Public Opinion 99 Table 4. Continued Democratic House Vote Democratic House Vote Democratic Senate Vote Democratic Governor Vote b p b p b p b p Independent Variables (RSE) (RSE) (RSE) (RSE) (.066) (.067) (.145) (.168) Black (.100) (.100) (.193) (.087) Male (.043) (.043) (.067) (.052) Gay marriage most important issue (.083) (.083) (.105) (.097) Iraq most important issue (.071) (.071) (.065) (.072) Terrorism most important issue (.089) (.089) (.105) (.127) Economy most important issue (.075) (.073) (.091) (.104) Health care most important issue Retrospective evaluation of national economy Retrospective evaluation of state economy (.070) (.069) (.063) (.100) (.028) (.028) (.062) (.054) (.047) (.045) (.074) (.118) Constant (.386) (.400) (1.126) (1.122) Pseudo R Log likelihood N 18,154 18,154 13,141 14,443 degree), income (a 12-point ordinal scale from less than $10,000 to $ ,000), marital status (married = 1, 0 otherwise), and race/ethnicity (binary variables for Black, Asian, and Latino respondents). DISPELLING MICRO- AND MACRO-LEVEL CONCERNS OF POSSIBLE ENDOGENEITY Before presenting our findings, we address the possibility that any effects that the minimum-wage ballot measures might have on opinions may be due to endogeneity. As should be clear, our modeling strategy overtly treats the presence of a minimum-wage initiative on a state s 2006 ballot as exoge-

16 100 Smith and Tolbert nous to public opinion, as well as exogenous to a state s economic conditions. Although the minimum-wage initiatives were not distributed randomly across the states, 15 there is compelling micro-level evidence that voters in the states where the measures qualified for the ballot were not a priori primed to think about the economy. Pre-election polls indicated that voters in the six minimum-wage states had opinions no different than those in the other 44 states with respect to raising the minimum wage. A Pew Research Center national poll of 1,405 adults conducted in March 2006 (seven months before our survey was in the field) found that 83 percent of the American public supported raising the federal minimum wage by $2.00 to $7.15 an hour, with 49 percent strongly supporting an increase. 16 A crosstab of support for increasing the minimum wage with respondents residing in the six states that would eventually have minimum wage on their November ballots reveals that percent of the respondents (n = 154) said they wanted the federal minimum wage increased. The level of support for raising the minimum wage among respondents residing in the other 44 states (n = 1,212) was nearly identical, at percent (Pearson chi 2 =.0108, p =.917). Subsamples by partisan groups reveal the same null findings. Pre-election polls suggest no evidence of endogeneity that might bias our results. At the macro-level, there is also no evidence that the possibility of disparate economic conditions across the states may have predisposed residents of the six initiative states to say the economy was the most important issue in the upcoming midterm election. With respect to the 2006 unemployment levels across the states, 17 a two-sample difference of means t-test reveals that average unemployment in the six states with the November ballot measures was 4.41 percent, which was not significantly different from the 4.35-percent mean unemployment rate of the other 44 states (t (48) =.129, p <.898). Similarly, differences in the mean hourly wage in 2006 across the two sets of states were 15. Although proponents of the minimum-wage initiatives considered voter preferences regarding minimum wage in the six states prior to their efforts to qualify the measures, they also considered voter preferences on minimum wage in other states. Proponents were well on the way to qualifying minimum-wage measures in Arkansas and Michigan, for example, when they voluntarily suspended their signature-gathering efforts after brokering deals with state lawmakers to increase the minimum wage through the traditional legislative process. See Arkansas Poised to Follow Michigan, Give Working Families a Raise, Brennan Center for Justice Press Release, April 3, Available at: michigan_give_working_families_a_raise/. 16. Michael Dimock, Maximum Support for Raising the Minimum: Most Americans Now Live in States That Have Raised the Wage Floor, Pew Research Center for the People & the Press, April 19, Available at: Response rates for the Pew surveys are 37 percent for 2006 (AAPOR Response Rate 3). 17. Data for the macro-level analyses are from the U.S. Bureau of Labor Statistics, Unemployment Rates for States, 2006, and U.S. Bureau of Labor Statistics, Occupational Employment and Wages, May 2006, pub_2006.htm.

17 Direct Democracy and Public Opinion 101 not statistically significant (t (48) =.541, p <.591), as the mean hourly wage in the six initiative states was $17.47, compared to $17.99 in the other 44 states. Finally, comparing the differences in the 2006 mean hourly wage of the lowest decile of the six initiative states workforce ($7.22) versus the other 44 states ($7.45) reveals no significant difference (t (48) =.757, p <.453). Thus, there appears to be no compelling evidence at the macro-level that the economic characteristics of the six states with minimum-wage measures differed significantly from the other 44 states. FINDINGS: WHO SUPPORTS INCREASING THE MINIMUM WAGE? Table 1 models the October 2006 CCES data using subsamples by partisanship with Democrats, Republicans, and Independents in columns 1, 2, and 3, respectively. Since the dependent variable is binary favor increasing the minimum wage or not logistic regression is used. Holding all other factors constant, Democrats living in a state with minimum wage on the ballot were more likely to support increasing the minimum wage than Democrats living in states without such ballot measures. Although we do not have panel survey data, this suggests that exposure to the ballot measure campaigns had the effect of modifying state-specific public opinion on this policy issue among a Democratic partisan subsample, especially when compared to the March 2006 Pew survey data. In contrast, as column 2 shows, Republicans residing in states with the ballot initiatives were less likely to favor increasing the minimum wage, and there was no difference in opinions on the policy among Independents living in a ballot state or not (column 3). It should be noted that all six states with minimum-wage measures on the ballot in 2006 voted for the Republican presidential candidate in 2004 and were not liberal or blue states per se. Table 2 repeats this analysis but instead uses the full sample of respondents and employs an interaction term (Democratic partisan residing in a state with minimum wage on the ballot). The coefficient for the interaction term is positive and statistically significant, meaning that Democrats residing in states with minimum wage on the ballot were more likely to favor increasing wages than were Democrats residing in states without ballot measures. This is consistent with the findings from the subsample analysis reported in table 1, column 1. The substantive magnitude of this effect is moderately strong. Holding all other variables in the model at their mean or modal values, simulations show that Democrats residing in states with minimum wage on the ballot are predicted to be four percentage points more likely to support increasing the federal minimum wage than Democrats residing in other states (s.d.=.008). These data provide evidence that direct democracy campaigns can help set the policy agenda in elections, consistent with previous work on gay marriage in the 2004 elections (Donovan, Tolbert, and Smith 2008) or nuclear freeze or education (Nicholson 2005, 2008). These agenda-setting effects are state spe-

18 102 Smith and Tolbert cific, appearing in states with measures on the ballot, and are not national. Here we go further, showing that exposure to direct democracy campaigns can temporarily shape public opinion on specific policy issues among a targeted partisan population (Democrats). Our findings are consistent with research suggesting that direct democracy can prime partisans to behave in certain ways (Branton 2003; Bowler and Donovan 1998; Smith and Tolbert 2001). FINDINGS: WHO THINKS THE ECONOMY IS THE MOST IMPORTANT ISSUE IN THE ELECTION? As we have noted above, the 2006 midterm elections have been generally considered to be a referendum on the Iraq War and President Bush s handling of the war. The outcome of the election led to the Democratic Party regaining control of the U.S. House and U.S. Senate. As mentioned earlier, compared to terrorism and Iraq, the economy and jobs in 2006 ranked as a relatively less salient issue (see discussion above). In the CCES post-election survey, 2,216 individuals (roughly six percent) of the 36,500 respondents named economy and jobs as the most important issue in the election. Notwithstanding the possibility of response bias that is inherent in any election survey, did residing in a state with minimum-wage ballot measure campaigns increase attention to the economy, especially among Democrats? Column 1 of table 3 indicates that Democrats living in states with minimum wage on the ballot were not only more likely to favor increasing minimum wage, but also more likely to think the economy was the most important issue in the election. The direct democracy campaigns appear to have primed Democrats to focus on the economy in these states. The initiatives appear to have increased the saliency of the economy in the 2006 election among Democratic partisans, which was one of the goals of the labor and progressive backers of the campaigns. These data provide some evidence that their policy priming strategy worked. But were Democrats residing in states with ballot measures also more concerned with the saliency of the Iraq War, terrorism, or other issues in the election? Columns 2, 3, and 4 of table 3 provide a counterfactual test. As the results show, Democrats residing in states with minimum-wage ballot measures were not more likely to think that Iraq, terrorism, or health care and health-care costs was the most important issue. The only issue which became more salient among Democrats with the direct democracy campaigns was the economy. This provides direct evidence that ballot measures can shape broader state-level issue publics. We also find that individuals who held negative evaluations of either their state s or the national economy and who were exposed to the ballot measure to raise the minimum wage (i.e., were primed to think about the economy) were more likely to say the economy and jobs was the most important issue in the election, compared with those who held negative economic evaluations and resided in states without a minimum-wage ballot measure (see Appendix).

19 Direct Democracy and Public Opinion 103 FINDINGS: DID MINIMUM-WAGE BALLOT MEASURES PRIME DEMOCRATIC CANDIDATE CHOICE? Table 4 approximates the key model presented in Donovan, Tolbert, and Smith (2008) using exposure to minimum-wage ballot measures instead of gay marriage, and 2006 survey data instead of 2004 data. Compared to previous research using Pew (Donovan, Tolbert, and Smith 2008) or NES data (Nicholson 2005) with smaller samples, our analysis draws on CCES data, with large samples within states. We are thus on firmer ground in evaluating the effect of state contextual effects on individual behavior than earlier work in this area. The dependent or outcome variables measure vote intentions for the Democratic candidate for U.S. House, U.S. Senate, and governor in a respondent s state. 18 As expected, column 1 indicates that if an individual favored increasing the minimum wage, he or she was more likely to vote for a Democratic congressional candidate, holding constant partisanship, union membership, ideology, socioeconomic status, policy issue positions, and economic evaluations. Column 2 introduces our ballot priming interaction, measuring whether the individual favored increasing the minimum wage and resided in a state with minimum wage on the ballot. The coefficients for a Democratic congressional and gubernatorial vote are statistically significant and positive (see columns 2 and 4). Residing in a state with an initiative campaign may have primed support for Democratic candidates over and above the effect of public opinion on the issue, holding all else constant. The substantive magnitude of these effects is considerable. Holding all other variables in the model in column 4 at their mean or modal values, a typical respondent who does not favor increasing the minimum wage and who does not reside in a state with the measure on the ballot has only a.25 probability of voting for a Democratic candidate for governor. When repeating the simulation for a respondent who favors increasing the minimum wage but does not reside in a ballot state, the probability of voting for the Democratic governor increases to.43. If a respondent favors the minimum wage and resides in a ballot state, the probability of voting for the Democratic governor jumps to.61, an increase of.18 based solely on exposure to the ballot measure. This may have been large enough to swing close candidate elections in these six states. Although the substantive effect on the probability of voting for a Democratic House candidate is roughly half that as for a Democratic governor, the minimum-wage initiatives still may have contributed to the Democratic Party s takeover of Congress in The larger impact of 18. We add to this model two binary variables for ideology (conservative and liberal), with moderates as the reference category. We do not control for whether any of the candidates took an explicit position on minimum wage, as there is no reason to assume that candidates (Republican, Democratic, or third party) in states without initiatives on the ballot would be asked to take a stance on this (non) issue. Following Nicholson (2005), we assume that voters rely on partisan stereotypes when evaluating candidates, regardless of a candidate s specific stance on minimum wage.