System Structure, Campaign Stimuli, and Voter Falloff in Runoff Primaries

Transcription

1 System Structure, Campaign Stimuli, and Voter Falloff in Runoff Primaries Charles S+ Bullock III The University of Georgia Ronald Keith Gaddie The University of Oklahoma Anders Ferrington The University of Oklahoma In a model that relies entirely on legal and structural factors, Wright (1989) seeks to account for changes in voter participation from the initial primary to the runoff. This article tests an alternative, campaign-centered model of voter turnout in 109 congressional runoffs from 1982 through The analysis indicates that candidate-centered factors, including the amount of money expended by the candidates in the runoff and the political experience of the primary leader, influence turnout in runoff primaries. Generally, when more money is spent during the runoff, voter participation declines less relative to the initial primary, suggesting that a more stimulated political environment encourages greater participation. Spending before the initial primary is less influential than spending between the primary and runoff in maintaining voter turnout, which indicates that any potential effects from stimulation of the environment in the prior campaign have largely dissipated by the time of the second election. The perceived health of a democracy is derived from the acceptance of and participation in the institutions of government. Social scientists bemoan the low rates of voter turnout even in presidential elections, a decline coincident with greater ballot access. Research on turnout has compared voter mobilization across voting districts in a single election, or aggregate turnout across elections, or examined the decision of individual voters to participate in a particular election, always focusing on a single instance of voting. This research note looks at the continued participation of voters in the majority-vote-with-runoff primary, a context that requires greater and more prolonged effort on the part of the voter. Data from 109 U.S. House runoffs from 1982 through 1996 are used to test a model of turnout in the runoff. An earlier version of this paper was presented at the annual meeting of the Midwest Political Science Association, April 27 30, 2000, Chicago. We thank Joe Aistrup and the editor for comments and suggestions. THE JOURNAL OF POLITICS, Vol+ 64, No+ 4, November 2002, Pp Southern Political Science Association

2 System Structure, Campaign Stimuli, and Voter Falloff 1211 Majority vote systems attract periodic interest of academics and the political classes, especially when third-party candidates prevent the emergence of a majority winner in the election (as happened in the U.S. presidential elections of 1992, 1996, and 2000) or when support for third-party candidates deny the election of a Condorcet method winner (as likely happened in the vote for Florida presidential electors in 2000). Despite the majoritarian attractiveness of the runoff election, the runoff system is beset by both perceived and real problems. A major substantive problem associated with the runoff is mobilizing the electorate for a second election. The double primary introduces interesting dynamics and constraints absent from one-shot elections. The beginning of the runoff campaign is clearly defined and the campaign is brief, typically no more than five weeks. The number of candidates is constant across all runoffs, unlike primaries and some general elections. Runoff candidates are aligned in direct opposition, and voters have no avenue for protest other than not voting. The vote in the initial primary provides voters with tangible evidence of candidate strength in place of the vagaries of polls. This prior performance means that there is a reduced information constraint on determining the odds of winning. 1 For the social scientist, runoffs provide a baseline of participation against which to measure subsequent interest. This particular institution also allows us to observe continued participation when voters are called upon to ratify or reconsider their initial choice and therefore imposes a rigorous test of the influence of both structural features and campaign effects on turnout. 2 In order to test the influence of campaign effects and the continuing utility of variables used by others, measures of campaign spending and candidate quality (campaigncentered variables not used previously) are included in our multivariate model. Voter Participation and Runoff Elections Majority vote requirements are used in a variety of electoral constituencies throughout the United States and abroad. In the United States, runoffs are most widely found in the South, where states adopted the majority vote requirement when bipartisan competition was rare and primary fields large. In 2000, more than one-fifth of U.S. House nominations were conducted using majority-vote rules (92 districts), and another 19 districts were affected by either a modified 1 Prior research on runoff elections identified a variety of myths associated with the majorityvote requirement, including the well-known leader loses myth. This myth is not borne out as primary leaders win runoffs about 70% of the time (Bullock and Johnson 1992). 2 The runoff creates an ideal venue for consideration of rational perspectives on voter participation. Kanazawa (2000) argues that voters act in the role of reinforcer or punisher, which substantially raises the individual perception of casting a decisive vote. This behavior (stochastic learning) is especially evident in elections that feature only two candidates and therefore no viable avenue for protest.

3 1212 Charles S. Bullock III, Ronald Keith Gaddie, and Anders Ferrington plurality-threshold rule (North Carolina) or a majority-vote general election rule (Louisiana). 3 Many cities employ a vote threshold in their electoral systems. The runoff is designed to nominate candidates who enjoy broad support. In a single-shot, plurality primary, a fringe candidate with a core of dedicated followers could win a small plurality in a crowded field (Key 1949). Requiring a majority for nomination forces candidates who survive to the runoff to broaden their coalitions and perhaps moderate their views. The experience of Arkansas, one of the last states to adopt the majority-vote requirement, is illuminating. It instituted a dual primary in an effort to prevent Ku Klux Klan candidates from winning office with only plurality support (Alexander 1944). The problem of Condorcet success is not entirely resolved by the runoff system. While the plurality system is least efficient in producing Condorcet winners, the runoff system confronts problems in producing a winner, especially if the Condorcet candidate is squeezed out of the second round by extremist candidates who run from either pole of the political spectrum (Merrill 1984). The runoff system does a satisfactory job of producing a Condorcet winner when policy voting is not the norm (Adams 1997). Another dimension worth considering, however, is whether the runoff system is an effective instrument for expressing the majority preference when the possibility of voter exit exists. How bad is the voter falloff problem in a runoff election, and what factors influence the magnitude of voter falloff? Turnout in the Runoff If costs are high relative to initial participation in a primary, then they will be higher in a double-primary where voters are called upon to vote twice for the same office(s) in a short period of time. Efforts in the short campaign leading up to a runoff have two goals: keeping supporters from the initial election mobilized and capturing voters who either voted for another candidate or stayed home at the initial primary. Efforts to maintain primary turnout levels have registered varying levels of success although participation generally declines. Key s (1949) examination of gubernatorial runoffs found that participation typically declined from the first to the second primary. More recent studies continue to find that participation rates drop between primaries and runoffs. Stephen Wright (1989) examined Democratic congressional and gubernatorial runoffs held between 1956 and 1984 and found that participation usually dropped from the primary to the runoff although runoff participation varied among states and across offices. Bullock and Johnson (1992), who studied runoffs held between 1970 and 1986 and added state legislative and executive contests to the types of elections studied by Wright, concur that participation usually drops. A study of North Carolina congressional and gubernatorial runoffs found that participation declined in 18 of 23 cases (Lanier 1983). However, Bullock s (1990) 3 In addition to a majority vote primary rule, Georgia has a 45% general election requirement.

4 System Structure, Campaign Stimuli, and Voter Falloff 1213 study of runoffs in six cities found that runoff participation usually exceeded turnout in the first round of voting. An analysis of Georgia county-level elections reported by Bullock and Johnson (1992) found that participation usually increased slightly in runoffs to choose nominees for sheriff, which is often the most important office in a rural county, but generally decreased in runoffs for school boards and county commissions. Previous efforts to explain runoff turnout have utilized legal and structural factors. While these will be included in our model, we introduce campaign finance and candidate experience, thus injecting two political variables into the mix. Campaign Effects Anthony Downs (1957) and William Riker (1982) argue that lack of information is a barrier to voter participation. Candidate advertising can help overcome that barrier. When more money is spent, more information gets to voters, which in turn stimulates turnout (Leighley and Nagler 1992). Patterson and Caldeira (1983) suggest that to understand the turnout problem, scholars must pay more attention to political mobilization, especially financing and candidate effects. They observe greater voter turnout in high-spending gubernatorial elections, despite controls for competitiveness. Cox and Munger (1989) observe that spending in competitive U.S. House general elections was associated with higher levels of turnout when a variety of structural and contextual factors were held constant. We expect to see a similar effect on continued voter participation in runoff primaries. We assume that money spent during the campaign is a surrogate for information dissemination that will stimulate participation. Since candidates can rarely buy advertising time on credit, measures of spending should be fairly accurate indicators of direct and immediate campaign effort in the short campaign. Most campaign spending in congressional contests is funneled into media buys while a not insignificant amount is devoted to voter contacting and mobilization (Weilhouwer and Lockerbie 1994). Mobilizing activities, including telephone banking, face-to-face canvassing, and direct mail, have variable influence on voter participation. The most effective of these techniques face-to-face canvassing and direct mail (Gerber and Green 2000, 2001) are costly in time, personnel, production, and mailing costs. The goal of the election is to win votes; spending in a sudden-death situation like the runoff presumably will be directed at capturing votes. We argue that money translates into information dissemination. In a second primary, voters come to the polls with potentially more information because of the extended campaign, a narrower range of choices and information derived from the first primary. Going into a runoff, voters should have more reliable information than in a one-shot election. Fey (1997) finds that preelection polls provide voters with information about the viability of candi-

5 1214 Charles S. Bullock III, Ronald Keith Gaddie, and Anders Ferrington dates. The information on candidate strength as demonstrated in a primary is more concrete than poll data. Although it may become dated, the results from the primary provide insights missing in the many congressional contests not probed by independent polls. Extensive research demonstrates that candidate attributes influence election outcomes. Politically experienced candidates are more likely to win open-seat nominations and elections (Bond, Fleisher, and Talbert 1997; Gaddie and Bullock 2000) and more likely to raise larger amounts of campaign cash (Herrnson 2000; Jacobson 1997). And while the odds of defeating incumbents are poor, challengers who have greater political experience are more successful than are novices (Jacobson 1997). The heightened name recognition that often accompanies political experience may stimulate greater runoff turnout since voters who know nothing about candidates and do not even recognize their names are less likely to trek back to the polls. Legal and Structural Effects Competitiveness has been identified as a stimulant to heightened participation (Cox and Munger 1989; Key 1949). If, in the initial primary, the frontrunner has a commanding lead over the runner-up, some voters may be disinclined to return to the runoff, assuming that the plurality leader will coast to victory. On the other hand, a narrow margin for the primary leader may trigger greater interest and participation in the second balloting. Bullock and Johnson s (1992) multivariate model indicated a negative but insignificant relationship between the magnitude of the frontrunner s margin in the primary and turnout in the runoff. Wright (1989) reaches the same conclusion using a different measure of primary competitiveness. Key (1949, 319) speculated that if one assumes that the decision to vote is predicated at least in part on the benefits derived from the act of voting, then more people should show up for a runoff if they have the opportunity to vote in multiple contests. Congressional runoffs held concurrent with statewide runoffs in the same party have less voter drop-off from primary to runoff (Lanier 1983; Wright 1989), and Bullock and Johnson (1992) report that turnout in the runoff was especially high in years when gubernatorial candidates were chosen. Bullock (1990), who frequently found high rates of participation in municipal council runoffs, examined only those in which a citywide office was being resolved in the runoff. All of the research points toward higher participation when the runoff ballot includes contests for higher offices. 4 4 The presence of multiple voting opportunities in one election reinforces turnout through campaign visibility and increasing the salience in an election to large numbers of voters. This thesis goes back at least as far as Angus Campbell s (1966) study of surge and decline, and is repeatedly reiterated in studies of voter mobilization in presidential and congressional elections (James Campbell 1986). More recent research shows that initiatives and referenda are related in increased voter turnout in major elections by activating voter sentiment on highly salient issues (Smith 2001).

6 System Structure, Campaign Stimuli, and Voter Falloff 1215 The continuous action hypothesis suggests that shorter time between a primary and a runoff provides less opportunity for voters to become distracted from the campaign (Wright 1989). The runoff becomes a seamless continuation of the initial primary so that candidates do not slacken efforts at voter mobilization even momentarily. On the other hand, voters may experience fatigue and not return to the polls when two contests are in close proximity. The voter fatigue hypothesis suggests that runoff participation will be stronger when a longer recovery time exists between the two rounds of balloting. With the GOP having become competitive only during the last generation in much of the South, the tradition of voting in runoffs held by that party is less well established than among Democrats. In areas of the South where the GOP is still emerging, it may have fewer contests resolved in a runoff, and with fewer contests there will be fewer reasons to vote. Bullock and Johnson (1992) show that participation was significantly less in Republican than Democratic runoffs. Lanier (1983) reports a drop-off rate of 89% in a Republican U.S. Senate runoff. We do not know how many down-ticket runoffs accompanied the congressional contests that are the subject of this research, so party may be a surrogate for the number of contests remaining to be resolved or for traditions of runoff voting. The dispersion of the vote in the initial primary across candidates may affect drop-off. Wright (1989) suggests that drop-off will be greater when candidates not making it into the runoff got larger shares of the vote in the initial primary because that may indicate a sizable component of the electorate for whom neither choice in the second round is attractive. Another possibility is that when the two candidates who advance to the runoff get almost all of the vote, it means that they were closely competitive since had either attained a majority, a runoff would have been unnecessary. Wright (1989) and Bullock and Johnson (1992) found stronger runoff turnout following primaries in which the share of the vote going to the candidates who went on to the runoff was greater. 5 Elections that are determinative are more likely to attract voters than are elections that only narrow the options. A runoff that precedes a general election in which only one party has a nominee might be perceived as more important than a runoff followed by a seriously contested general election. Inter-party competition increases general election turnout (Cox 1999; Hoffstetter 1973), and Gray (1976) observed evidence that Key s (1949) competitive threat hypothesis that is, party leaders will try harder to mobilize an electorate in a competitive party environment was empirically borne out primarily in the South. With partisan competition becoming more widespread in the states that rely on runoffs, the second primary may have come to be viewed as less significant resulting in a drop in participation. Wright (1989) demonstrated that as the share of the vote won by the Republican nominee in the general election increased, participation in the Democratic runoff declined. 5 Similar results have been reported for French runoffs (Pierce 1981).

7 1216 Charles S. Bullock III, Ronald Keith Gaddie, and Anders Ferrington Finally, Wright (1989) found that open primaries result in lower runoff turnout. We expect to see this finding repeated in our analysis. Data and Method Hypotheses about voter turnout under the majority vote rule are tested using 109 congressional runoff elections held in nine southern states between 1982 and The quality and consistency of campaign finance data fall off precipitously before 1982, precluding analyses before then. Hypotheses on voter participation in runoff primaries are tested in a multiple regression model in which the dependent variable is voter participation in the runoff measured as a percentage of primary participation. Because the dependent variable is interval and unbounded, OLS regression is an appropriate technique. Turnout data come from the most recent edition of Congressional Quarterly s Guide to US Elections and various issues of Richard Scammon s America Votes series (Scammon ). The specification and measurement of the independent variables follows. Campaign Spending: Two measures of campaign spending are included in the models. The first is the total amount spent in the initial primary by all candidates; the second is the money spent in the period between the initial primary and the runoff by the remaining two candidates. 7 The figures were calculated from Federal Election Commission candidate fund-raising reports and standardized to constant 1994 dollars. To capture potential declining returns from spending as expenditures grow, the multivariate models are estimated using a linear measure of spending and a quadratic of spending as recommended by Patterson and Caldeira (1983), Cox and Munger (1989), and Grier (1989). We expect that the bulk of the spending impact will come from spending in the runoff. Candidate Quality: A pair of dummy variables are coded 1 if a respective candidate held elective office, 0 otherwise. Data on candidate experience for both the initial primary front-runner and runner-up come from various issues of CQ Weekly Report. Competitiveness in the Primary: Primary competitiveness is measured as the margin of the first-place finisher over the runner-up in percentage points. Other Statewide Primaries: The presence of an up-ticket runoff for governor or senator in the same party is coded 1, and 0 otherwise. 6 Alabama, Arkansas, Florida, Georgia, Mississippi, Oklahoma, South Carolina, and Texas had majority vote requirements for congressional nominations during the period studied. North Carolina used a majority nomination requirement until 1989, when it adopted a 40% threshold for nomination. North Carolina races prior to 1989 are included in the analysis. 7 An alternative specification of the model treated the amount of money spent by the frontrunning and second-place candidates as separate variables. This specification did not provide as good a fit to the data, suggesting that it is the total stimulation of the electorate that matters, rather than the influence of spending by any particular candidate.

8 System Structure, Campaign Stimuli, and Voter Falloff 1217 Time to Runoff: The time period between initial primaries and runoffs ranges from one to five weeks. A variable to indicate the time-span was coded accordingly, and the quadratic of that value was included to check for a potential curvilinear relationship. Dispersion of the Vote: Dispersion of the primary vote is measured as the percentage of the votes captured by the first- and second-place finishers in the initial primary. Party of Primary: Republican primaries are indicated by a dichotomous variable coded 1 for GOP primaries, 0 otherwise. Partisan Competitiveness: Inter-party competition is measured as the share of the vote won in the general election by the nominee of the party not holding the runoff. Open/Closed Primary: Open primaries are coded 1, closed primaries 0. Analysis The decline in participation is far more widespread than in studies that examined earlier periods conducted when runoffs were more likely to be decisive in selecting officeholders. Of the 109 runoff elections examined, voter turnout increased from the first to the second primary in only four (3.7% of cases). 8 The average falloff in participation from the primary to the runoff amounted to almost a third of the initial primary electorate, 31.7% (see Table 1). The change in participation ranged from a 7% increase in voter turnout to a decline of almost 90%. To put this in perspective, Wright (1989) found runoff turnout averaged 86% of the Democratic primary vote with the second round drawing more voters 21% of the time in House contests ( ). Bullock and Johnson (1992) observed an average vote decline of 20% in House contests ( ) with one-sixth of the runoffs attracting more voters than the primary. Combining these two studies with our results shows increasing falloff over time. Change in voter participation is related to the total level of spending in the second primary. As indicated in Figure 1, the level of voter falloff is greater in districts where spending is less. 9 A loose, positive relationship is evident in both the Democratic and Republican primaries, though the pattern is more clearly pronounced in the Democratic cases. The multivariate model in Table 2 explains 45% of the variation in the change in voter turnout. Eight of the fourteen variables are statistically significant. 10 The impact of campaign spending between the two primaries is especially 8 The turnout change variable, though slightly skewed, is distributed in a relatively normal fashion around the mean. 9 The plotted data are of logged spending totals against the relative falloff in turnout. 10 Checks for multicollinearity revealed no problems.

9 1218 Charles S. Bullock III, Ronald Keith Gaddie, and Anders Ferrington TABLE 1 Descriptive Statistics Mean Sd % Cases % Change in Turnout from Primary to Runoff Total Expenditures, Primary Total Expenditures, Runoff Primary Leader Political Experience 31 Primary 2 nd Place Political Experience 35 Initial Primary Margin Weeks Between Primary and Runoff Leader 2 nd Place % in Primary GOP Primary 49 Party Competitiveness Other Statewide Primary 20 Open Primary 46 Note: For dichotomous nominal variables, the percent of cases coded 1 is indicated in the third column. pronounced on continued voter mobilization. The impact is nonlinear. If we assume $100,000 spent between the primary and the runoff, the net impact on the change in voter turnout is just 1.6 points; at $250,000 spent, the impact is an increase of 23.8 points; at $500,000, the impact is a net increase of 30.0 points, all other influences held constant. In a voting system that requires voters to turn out more than once, more campaign spending provides continued stimulation, and apparently encourages participation, up to a point. With runoff spending averaging less than $100,000, it does little to spur turnout in a number of contests. Spending substantially affects turnout in the 26 runoffs in which more than $150,000 was spent. Diminishing returns from spending begin at about $950,000, and further spending is linked to falling rates of participation. Candidate experience also has an impact, although only the experience of the front-runner is significant. Voter turnout in the runoff is almost 14 points higher if the initial primary leader is an officeholder. A politically experienced front-runner may stimulate greater runoff participation in several ways. First, the campaign organization that brought the front-runner success previously, along with the enthusiasm produced by leading the primary, can help turn out votes in the runoff. Second, the visibility that comes with holding office may bring additional people to the polls, with some wanting to vote for a known quantity while perhaps others seek to thwart the ambitions of a candidate whom they had opposed previously. This significant relationship may be of dubious value to experienced front-runners. It is not necessarily an advantage to the front-

10 System Structure, Campaign Stimuli, and Voter Falloff 1219 FIGURE 1 Spending and Change in Turnout from First Primary to Subsequent Runoff

11 1220 Charles S. Bullock III, Ronald Keith Gaddie, and Anders Ferrington TABLE 2 Change in Voter Turnout from Primary to Runoff b (s.e.) t Total Expenditures, 1 st Primary ($100,000) (.016) Total Expenditures, 1 st Primary ($100,000)squared (.0009) Total Expenditures, 2 nd Primary ($100,000) (.280) Total Expenditures, 2 nd Primary ($100,000)squared (.0037) Primary Leader Experience (3.65) Primary 2 nd Place Experience (4.34) Primary Vote % of 2 leading candidates (.14) Primary Margin (.21) Concurrent Statewide Runoff (3.69) Weeks since Primary (10.68) Weeks since Primary squared (1.55) Open Primary (3.93) GOP Primary (3.43) Party Competitiveness (.19) Intercept R-Square0 Adjusted R-Square N 109 Dependent Variable: Change in votes from primary to runoff as a percentage of primary ballots. Italicized t-statistics are significant at the.05 level using one-tailed tests. runner to have a high degree of electoral mobilization in the second primary. Previous research finds that the likelihood of a primary front-runner winning a runoff is related to lower turnout in the second election (Bullock, Gaddie and Ferrington 2001). To the extent that voter turnout in the second primary is a

12 System Structure, Campaign Stimuli, and Voter Falloff 1221 product of the front-runner s quality, it is a function of voters mobilizing against the experienced front-runner. 11 Reinforcing Wright (1989), the presence of a statewide runoff led to turnout 9.7 points higher than in runoffs unaccompanied by a gubernatorial or Senate contest. GOP primaries remain less likely to attract voters for a second round, despite the heightened prospects for GOP nominees since 1982 (cf Gaddie and Bullock 2000). Runoffs in open primary states experience more falloff than when participation is restricted to party loyalists, reinforcing Wright s (1989) findings. In conformity with earlier work (Bullock and Johnson 1992; Wright 1989), while the sign indicated runoff turnout might be higher in the wake of closely contested primaries the relationship was not statistically significant. The persistent failure to find a strong relationship between primary competitiveness and runoff turnout accords with the widely held leader loses myth. While the myth is accurate only about one-quarter of the time in House races (Bullock and Johnson 1992; Ewing 1953), voters who believe it may discount the size of the primary leader s advantage. Whether these voters participate in the runoff is more likely affected by post-primary campaign spending and other factors. The time between primary and runoff becomes significant when measured with a quadratic decay. Voter turnout falls off least in a mid-length runoff campaign of three weeks and is substantially less in runoffs that follow immediately on top of the first primary or that occur too long after the first primary. It appears that there is a fine line between voter fatigue in the short term and a loss of interest in participating in the long term that significantly affects voter interest in returning to the polls. If measures of campaign spending and candidate quality are removed, the explanatory power of the model drops by over two-thirds. The results of that analysis are: Change in Turnout Primary margin (.247) Statewide Primary (12.640) Weeks since primary ( 2.245) Lead and Second candidates %(.121) GOP Primary ( ) Party Competitiveness (.03062) Open Primary ( 6.149); Adjusted R-square.114. Only GOP Primary and Statewide Runoff are significant at the.05 level, clearly indicating the superiority of the campaign-based variables to explain the variation continued voter mobilization. These results, while weaker than Wright s (1989), are in more line with those of Bullock and Johnson (1992). Possible causes for the model performing less well than Wright s are the differences in time period (only two elections overlap with Wright), our exclusive focus on what was the least visible office in the Wright study, inclusion of runoffs in both parties, and a greater tendency for turnout to decline now than in the past. 11 This result may be an artifact of the nature of the races decided in runoffs. Many experienced candidates win nomination without a runoff. Those who are propelled into a second round have probably squandered their quality advantages and become a target for a successful voter mobilization in support of the underdog second-place finisher.

13 1222 Charles S. Bullock III, Ronald Keith Gaddie, and Anders Ferrington Conclusions Voter participation is a product of legal structures, the political environment, and specific and episodic campaign and candidate effects. This note examines voter participation under a system that exacts a higher cost on voters than most by requiring multiple elections to determine an outcome. The runoff impacts voter participation, as first-primary turnout typically falls by almost one-third. Continued voter participation is encouraged by the quality of the front-running candidate and by the stimulation of the campaign environment with money up to a point. The quadratic term for the time elapsed between the primary and the runoff indicates a curvilinear relationship and suggests fatigue immediately following a primary and disinterest several weeks out from the primary. Potential turnout peaks when the delay is moderate in length (around three weeks). Campaign experience and spending tap elements unique to each campaign that have not been used in other studies of runoff participation. While this note does not replicate in entirety the work of previous scholars, it is notable that several variables tapping context and previously found to be important were not useful here. Variables found to be important by earlier works but not here include the primary vote share won by the candidates who advanced to the runoff (Bullock and Johnson 1992; Wright 1989) and party competitiveness (Wright). These findings indicate that maintenance of voter interest in a double election system is greater when sufficient campaign activity takes place over a reasonable period of time. The nature of the campaign for nomination in the runoff system reflects the relative importance of campaigns over structure in the contemporary political system and the role of money as an engine for mobilizing and remobilizing the electorate. There is a limit to the utility of stimulating a political environment. The quadratic specification of spending shows that the highest turnout does not occur when spending levels are maximized. Moreover, there appears to be little carryover to the second primary from the spending in the initial primary, leading to the conclusion that the effects of the initial primary are confined to stimulating voters in that first election. The runoff primary becomes a new game, a new campaign, affected largely by the factors endogenous to that event. The structural influences play a secondary role in the game. Manuscript submitted 3 April 2001 Final manuscript received 12 March 2002 References Adams, James Condorcet Efficiency and the Behavioral Model of the Vote. Journal of Politics 59(4): Alexander, Henry M The Double Primary. Arkansas Historical Quarterly 3:

14 System Structure, Campaign Stimuli, and Voter Falloff 1223 Bond, Jon R., Richard Fleisher, and Jeffrey C. Talbert The Experience Factor in Open Seat Congressional Elections, Political Research Quarterly 50(2): Bullock, Charles S. III Turnout in Municipal Elections. Policy Studies Review 9(1): Bullock, Charles S. III, and Loch K. Johnson Runoff Elections in the United States. Chapel Hill: University of North Carolina Press. Bullock, Charles S. III, Ronald Keith Gaddie, and Anders Ferrington When Experience Fails: The Experience Factor in Congressional Runoffs. Legislative Studies Quarterly 26(1): Campbell, Angus Surge and Decline: A Study of Electoral Change. In Elections in the Campaign Order, eds. Angus Campbell, Philip E. Converse, Warren E. Miller, and Donald E. Stokes. New York: Wiley. Campbell, James E Presidential Coattails and Midterm Losses in State Legislative Elections. American Political Science Review 80(1): Cox, Gary W Electoral Rules and the Calculus Mobilization. Legislative Studies Quarterly 24(3): Cox, Gary W., and Michael C. Munger Closeness, Expenditures, and Turnout in the 1982 U.S. House Elections. American Political Science Review 83(1): Downs, Anthony An Economic Theory of Democracy. New York: Harper Collins. Ewing, Cortez A. M Primary Elections in the South. Norman: University of Oklahoma Press. Fey, Mark Stability and Coordination in Duverger s Law: A Formal Model of Preelection Polls and Strategic Voting. American Political Science Review 91(1): Gaddie, Ronald Keith, and Charles S. Bullock III Elections to Open Seats in the U.S. House: Where the Action Is. Lanham, MD: Rowman and Littlefield. Gerber, Alan S., and Donald Philip Green The Effects of Canvassing, Telephone Calls, and Direct Mail on Voter Turnout: A Field Experiment. American Political Science Review 94(3): Gerber, Alan S., and Donald Philip Green Do Phone Calls Increase Voter Turnout? A Field Experiment. Public Opinion Quarterly 65(1): Gray, Virginia A Note on Competition and Turnout in the American States. Journal of Politics 38(1): Grier, Kevin B Campaign Spending and Senate Elections, Public Choice 58(2): Herrnson, Paul S Campaigning for Congress. Washington DC: Congressional Quarterly Press. Hoffstetter, C. Richard Inter-Party Competition and Electoral Turnout: The Case of Indiana. American Journal of Political Science 17(2): Jacobson, Gary C The Politics of Congressional Elections. Boulder, CO: Longman. Kanazawa, S A New Solution to the Collective Action Problem: The Paradox of Voter Turnout. American Sociological Review 65(2): Key, V. O Southern Politics: In State and Nation. New York: Knopf. Lanier, Mark The Runoff Primary: A Path to Victory. N. C. Insight (June): Leighley, Jan E., and Jonathan Nagler Individual and Systemic Influences on Turnout: Who Votes? Journal of Politics 54(1): Merrill, Samuel III A Comparison of Efficiency in Multicandidate Electoral Systems. American Journal of Political Science 28(1): Patterson, Samuel C., and Gregory A. Caldeira Getting Out the Vote: Participation in Gubernatorial Elections. American Political Science Review 77(3): Pierce, Roy Left-Right Perceptions, Partisan Preferences, Election Participation, and Partisan Choice in France. Political Behavior 3(1): Riker, William H The Two-Party System and Duverger s Law: An Essay on the History of Political Science. American Political Science Review 76(4):

15 1224 Charles S. Bullock III, Ronald Keith Gaddie, and Anders Ferrington Scammon, Richard, ed America Votes. Washington D.C.: Elections Research Center, Congressional Quarterly. Smith, Mark A The Contingent Effect of Ballot Initiatives and Candidate Races on Turnout. American Journal of Political Science 45(3): Weilhouwer, Peter W., and Brad Lockerbie Party Contacting and Political Participation, American Journal of Political Science 38(1): Wright, Stephen G Voter Turnout in Runoff Elections. Journal of Politics 51(2): Charles S. Bullock III is professor of political science, University of Georgia, Athens, GA Ronald Keith Gaddie is associate professor of political science, University of Oklahoma, Norman, OK Anders Ferrington is a Ph.D. candidate in political science, University of Oklahoma, Norman, OK