[ Pobierz całość w formacie PDF ]
.555Age.00 (.00).007Hispanic.04 (.03).208Black.02 (.04).565Asian.20 (.04).000Education.03 (.01).060Income.02 (.01).027Constant.22 (.09).014N 1414F 67.83.000R2.31Note: Unstandardized instrumental variable regression coefficients, with robust standard errors in paren-theses to correct for heteroskedasticity.Reported probabilities based on two-tailed test.Second-stage estimates.First-stage estimate reported in column 2 of table 3.1.The instrumental variable in the first stage is totalexpenditures on ballot initiatives in the respondent s state.Source: Postelection Pew Internet and American Life Daily Tracking Survey (November 2002).predictor of voting in the 2002 election.Initiative campaigns (measured bytotal expenditures on ballot initiatives) appear to generate political interest,and the increased attention to the election significantly increases theprobability of voting, as expected.Again, the demographic control vari-ables are in the expected direction, as the older, wealthier, and more edu-cated were significantly more likely to vote, as were those who regularlyconsumed newsprint media.Two-stage causal models suggest that ballotmeasures increase voter turnout by stimulating interest in elections.Thefindings are consistent with previous research (Smith 2001; Lacey 2005;Smith and Tolbert 2004) that finds that states with salient initiatives orreferenda on the ballot have higher voter turnout over time; our surveyanalysis helps explain why we see a turnout effect.Ballot propositions mayhelp engage citizens in low-information elections, thus creating a positiveeffect on turnout.PPL-US_DDI-Bowler_CH003.qxd 2/27/2008 10:29 PM Page 4848 CAROLINE TOLBERT AND DANIEL BOWEN9.Presidential ElectionAs a check on the robustness of the findings and to compare differencesbetween midterm and presidential elections, we draw on the unique 2004preelection Pew survey data discussed earlier to predict general interest inelection.16 The dependent variable is identical to the one used for the 2002election data analysis: a binary variable measuring how much thought theindividual has given to the upcoming presidential election.17 Again, twoidentical explanatory variables are used: (1) the number of initiativesappearing on the respondent s statewide ballot in the 2004 election and (2)the total expenditures on ballot initiatives in the respondent s states.Valueson the number of initiatives ranged from 0 in noninitiative states to a highof 8 in California.Total expenditures on initiatives ranged from approxi-mately $10,000 in North Dakota to $202 million in California.State con-textual controls measure the competitiveness of senate, governor, andpresidential races in 2004 by the vote margin (one-vote margin), withhigher values indicating more competitive races.We expect citizens resid-ing in states with more competitive candidate races to be more interestedin the election.The model also includes standard ideological and demo-graphic control variables, chosen to mirror those used in the 2002 surveyanalysis.Democratic identifiers and Republican identifiers are coded 1,and independents and others coded 0.The respondent s level of educa-tional attainment,18 age (estimated in years), gender (with a binary vari-able coded 1 for males), race (with a binary variable coded 1 for AfricanAmericans, Latinos, and Asians, with whites as the reference group), andincome are accounted for.19Consistent with previous research showing a more limited effect ofdirect democracy in presidential elections, exposure to ballot initiativesdoes not seem to increase civic engagement in high-information presiden-tial elections (see table 3.4, column 1).20 This may explain the previousfinding that ballot initiatives in presidential elections have about half theturnout effect that they do in midterm elections (Tolbert and Smith 2005).The covariate for the number of initiatives on an individual s statewideballot in 2004 is not a statistically significant predictor of general politicalinterest.Similarly, higher total expenditures on ballot initiatives are notassociated with increased political interest in 2004 (column 2, table 3.4).Even the more subtle and direct measure of issue-campaign activity doesnot increase interest in the election for the overall population.However, even in high-information presidential elections, ballot initiativesand their associated campaigns appear to boost engagement among the loweducated.The same interaction models used in table 3.1 are reported in table3.4: respondent s education * number of initiatives on the ballot (column 3)Table 3.4 Impact of exposure to ballot initiatives and initiative campaigns on political interest, 2004 presidential electionInterested in politics 2004Model 1 Model 2 Model 3 Model 4b(s.e.) p z b(s.e.) p z b(s.e.) p z b(s.e.) p zNumber of initiatives on state ballot.02 (.02).500.14 (.07).046Total expenditures on initiatives.00 (.00).257.01 (.00).000Number of initiatives * education.03 (.015).057Total expenditures * education.00 (.00).000Newspaper news consumption.62 (.30).038.62 (.30).037.63 (.30).035.63 (.30).035Television news consumption.30 (.22).173.30 (.22).171.30 (.22).170.31 (.22).167Online news consumption 1.16 (.44).008 1.16 (.44).008 1.19 (.44).007 1.20 (.44).006Republican.52 (.20).008.52 (.20).008.52 (.20).008.52 (.20).008Democrat.52 (.17).002.52 (.17).002.52 (.17).002.53 (.17).002Male.03 (.15).826.03 (.15).824.04 (.15).798.04 (.15).796Age.01 (.00).002.013 (.00).002.01 (.00).002.01 (.00).002Hispanic.60 (.33).069.61 (.33).067.62 (34).068.62 (.34).066Black.37 (.26).155.37 (.26).157.38 (.26).154.37 (.26).16Asian.45 (.65).486.46 (.64).471.42 (.65).517.42 (.64).517Education.30 (.05).000.30 (.05).000.34 (.05).000.34 (.05).000Income.07 (.03).032.07 (.03).032.07 (.03).037.07 (.03).000Attend church.05 (.05).274.05 (.05).275.05 (.05).264.05 (.05).257Competitive election president.86 (.98).380.91 (.94).331.81 (.98).413.86 (.94).360Competitive election governor.04 (.20).841.06 (.21).790.04 (.21).852.06 (21).787Competitive election senate.07 (.21).749.06 (.20).744.07 (.21).733.07 (.20).737Constant 2.68 (1.01).008 2.72 (.98).006 2.82 (.99).004 2.87 (.96).003N 1284 1284 1284 1284Wald Chi2 172.28.000 168.74.000 187.29.000 447.53.000Pseudo R2.09.09.10 [ Pobierz całość w formacie PDF ]