Explaining Major and Third Party Candidate Entry in U.S. Gubernatorial Elections, 1980–2005 Conor M. Dowling, Yale University Steve B. Lem, Kutztown University

abstract While much scholarly research has focused on the quality of candidates in U.S. elections, the quantity of candidate participation has received much less attention. Consequently, there is little systematic evidence on what factors stimulate or dampen candidate entry into electoral contests, especially at the state level. In this article, we examine the effects of electoral laws, states’ social and political climates, and incumbency on gubernatorial candidacies. We test our claims using an event-count model with the number of candidates contesting each gubernatorial election from 1980 to 2005 as the outcome of interest. We divide total candidate participation into major and third party candidates to observe whether the factors affect neither, one, or both types of candidacies. We find that electoral laws and incumbency play an important role in determining the number of candidacies in these elections, but the effect of these factors varies for the two types of candidates.

elections are the lifeblood of democracy. But for elections to accomplish their requisite function—connecting the preferences of the electorate to governing officials—they must offer voters a choice. Without choices, the expressive act of voting has no meaning. While much scholarly research has focused on the quality of candidates in U.S. elections, the quantity of candidate participation—that is, the number of choices available in an election— has received much less attention. As a result, there is little systematic evidence concerning the factors that stimulate or dampen candidate entry into electoral contests, particularly in state and local elections (but, see Stratmann 2005 on the state legislative elections of 1998 and 2000; Bardwell 2002 and Lazarus 2008 on gubernatorial elections). The quantity of candidates entering the electoral arena is, of course, essential to understanding the dynamics of American democracy. As Duverger State Politics and Policy Quarterly, Vol. 9, No. 1 (Spring 2009): pp. 1–23 ©2009 by the Board of Trustees of the University of Illinois

2   dowling and lem

(1951) noted more than five decades ago, the single-member-district, plurality electoral system discourages third party candidate participation, while promoting competition between two major choices: Democrats and Republicans. The existence of two major parties, however, does not always guarantee choices on the electoral agenda. For example, the presence of a strong incumbent may deter would-be challengers from the opposing party, leading to an uncontested electoral race (see Squire 1989, 2000 on U.S. House and state legislative elections, respectively). Despite these conditions, substantial variation in the number of both major and third party candidates that enter electoral contests remains unexplained across the American states. In this article we ask: what accounts for these observed differences? We approach this question by developing a model based on three electoral characteristics. First, we focus on institutional differences with the expectation that electoral laws can significantly encourage or dampen the number of candidacies. Second, we analyze the social and political climates of the states because contextual differences can influence electoral participation. Last, we include election-specific factors, such as the presence of an incumbent. After deriving our hypotheses, we test our model using data for state gubernatorial elections. We find that institutional differences and election-specific factors affect gubernatorial candidacies in important and systematic ways.

explaining candidate entry in gubernatorial elections The average number of major and third party candidates in gubernatorial elections from 1980 to 2005 are presented by state in Table 1.1 There is a large range in the number of both major and third party candidacies. Specifically, the average number of major party candidacies ranges from 2.86 (Delaware and North Dakota) to 12.33 (California2). Similarly, the average number of third party candidacies ranges from 0.17 (Maryland and Nebraska) to as many as 6.5 (Minnesota). To explain this variation, we begin by developing a general model of candidate emergence based on Black’s (1972) decision-making calculus for a potential candidate, which is given by the following equation: U(O) = (PB)—C where, U(O) = the utility of the target office, O; P = the (potential) candidate’s estimate of his probability of winning the target office; B = the benefit the candidate receives in attaining the target office; and C = the cost required

Table 1.  Candidate Participation in Gubernatorial Elections, 1980–2005 State AL(6) AK(6) AZ(6) AR(8) CA(6) CO(6) CT(6) DE(7) FL(6) GA(6) HI(6) ID(6) IL(6) IN(7) IA(6) KS(6) KY(6) LA(6) ME(6) MD(6) MA(6) MI(6) MN(6) MS(6) MO(7) MT(7) NE(6) NV(6) NH(13) NJ(7) NM(6) NY(6) NC(7) ND(7) OH(6) OK(6) OR(6) PA(6) RI(10) SC(6) SD(6) TN(6) TX(6) UT(7) VT(13) VA(7) WA(7) WV(7) WI(6) WY(6)

Average Number of Major Party Standard Candidates Deviation State 8.50 8.17 5.67 5.75 12.33 4.17 3.67 2.86 8.17 7.83 9.5 4.67 5.67 3.86 5.00 7.33 7.83 10.33 6.33 6.00 4.83 5.83 4.17 7.17 7.71 6.14 8.33 11.83 6.00 8.71 7.67 3.67 7.86 2.86 3.50 8.12 10.17 5.33 3.70 4.83 4.33 9.50 7.17 4.29 3.38 3.29 9.43 10.43 4.67 6.50

2.07 1.94 1.97 2.12 4.37 2.04 1.86 0.90 5.56 3.54 1.05 1.97 1.63 0.69 0.90 2.58 2.79 2.88 4.08 2.10 0.75 2.71 3.66 1.83 3.30 2.04 3.88 3.45 1.53 6.10 1.03 1.03 2.8 1.07 1.87 3.71 2.93 3.33 1.34 2.23 2.34 2.35 4.07 1.70 1.26 0.95 3.78 6.08 1.75 2.81

AL(6) AK(6) AZ(6) AR(8) CA(6) CO(6) CT(6) DE(7) FL(6) GA(6) HI(6) ID(6) IL(6) IN(7) IA(6) KS(6) KY(6) LA(6) ME(6) MD(6) MA(6) MI(6) MN(6) MS(6) MO(7) MT(7) NE(6) NV(6) NH(13) NJ(7) NM(6) NY(6) NC(7) ND(7) OH(6) OK(6) OR(6) PA(6) RI(10) SC(6) SD(6) TN(6) TX(6) UT(7) VT(13) VA(7) WA(7) WV(7) WI(6) WY(6)

Note: Number of elections for each state reported in parentheses.

Average Number of Third Party Candidates

Standard Deviation

1.00 6.00 2.50 0.25 3.33 2.33 1.67 0.71 0.67 1.83 2.83 1.17 1.33 1.43 2.17 1.83 0.83 3.33 2.5 0.17 1.50 2.00 6.5 1.67 2.00 0.71 0.17 1.83 1.23 2.71 0.33 5.17 1.71 1.00 1.67 1.67 1.83 2.00 1.10 1.00 0.83 2.00 2.33 2.86 3.85 0.86 1.57 1.57 3.17 0.50

2.00 3.03 2.74 0.46 1.37 0.82 1.21 0.76 0.82 2.14 3.37 0.75 0.52 0.79 2.04 1.17 1.17 2.42 1.05 0.41 1.05 2.45 3.73 1.51 2.24 0.76 0.41 1.17 0.93 3.35 0.52 2.14 1.11 1.83 1.63 0.41 1.72 1.10 1.10 0.89 0.98 1.79 1.21 1.95 2.44 0.69 0.98 2.07 1.94 0.55

4   dowling and lem

to obtain the target office. When the value (B) of the target office (O) in combination with the probability (P) of victory is greater than the cost (C) of running for the target office, the utility (U) of running for office (O) is positive and the potential candidate declares candidacy (also, see Jacobson and Kernell 1981). State differences and electoral characteristics encourage or suppress candidacies by the degree that they influence the probability of winning (P), the benefit of office (B), and the cost of running for office (C). For example, ballot access requirements directly affect the cost of running, which should directly affect an individual’s willingness to declare a candidacy. Therefore, if one is willing to presume that the number of candidates in a particular election is simply an aggregation of individuals utilizing the decision-making calculus, we should be able to directly observe a correlation between our three types of electoral characteristics and the number of candidates contesting an election. Nevertheless, we need to make one important distinction before turning to a discussion of electoral characteristics and the number of gubernatorial candidates. Specifically, while the decision-making calculus provides a general framework to connect state and electoral characteristics to candidate entry, the factors we posit below might not have the same effect on the decisionmaking equation for every candidate. In this article, we distinguish between major party candidates and third party candidates (minor party candidates and independents), given that the resources needed to overcome some of the barriers to entry are typically more easily attainable by members of one of the two major parties.

Electoral Institutions and Candidate Entry As several scholars have noted (e.g., Duverger 1951; Cox 1997), the rules of the electoral game play a large role in shaping democratic participation. Unsurprisingly, then, one should expect variation in state electoral institutions to have an effect on an individual’s opportunity and willingness to contest public office. Ballot Access Requirements. One of the most straightforward ways that candidate entry can be restricted is by manipulating the requirements that must be met in order for a candidate to appear on the ballot.3 Specifically, state laws regulating ballot access may influence the total number of candidates that contest gubernatorial elections by affecting the costs of contestation (C). While all states institute ballot access restrictions in some form, we highlight two that are directly related to candidate entry: signature requirements and major party candidate filing fees.



spring 2009  /  state politics and policy quarterly   5

Both requirements are serious obstacles to major party candidate entry. At least one of these requirements is present in all but four states (Colorado, Connecticut, North Dakota, and Wyoming). Of the 46 states that have requirements, 25 require filing fees, 15 require petitioning with a list of signatures, and six require both. For third party candidates, signature requirements are present in every state save three (Florida, Louisiana, and Oklahoma). Therefore, as nearly every state has these requirements, the barrier to entry is the level of the requirement needed to appear on the ballot.4 For instance, some states require modest filing fees (e.g., $75 in Hawaii) for major party candidates, while others require substantial amounts of money to appear on the ballot (e.g., $4,000 in Texas). Clearly, the cross-state variations in these requirements affect the utility equations of the potential candidates for governor within each state. More specifically, as signature requirements and filing fees increase, the costs of running for office (C) increase, and consequently, the total number of (major and third party) candidates should decrease. Party Endorsement of Candidates. Another electoral law that is likely to play a role in the candidate entry process is whether a state allows for preprimary party endorsement. Bardwell (2002) found that gubernatorial incumbents are less likely to be contested in their primary if they are from a party-endorsement state. Similarly, we argue that states with preprimary party endorsement of candidates decrease the probability (P) of winning for those potential candidates that do not receive the party’s endorsement, ultimately resulting in fewer major party candidates.5

The Social and Political Climate of a State and Candidate Entry While formal state institutions play an important role in shaping candidate entry, a state’s social and political climate—the norms and patterns of behavior particular to each state—can also influence the number of individuals that ultimately contest an election. Hence, if there is a way to generalize about the social and political climate of states, rather than merely claim that each state is different in its own right, these generalizations can be included in a model of candidate entry. Accordingly, we consider two aspects of a state’s social and political climate: political competition and social capital. Competition. The degree to which the two major parties are competitive in a state is likely to have an effect on candidate participation.6 However, major party competition will influence major party and third party candidacies in different ways.

6   dowling and lem

We begin with the relatively straightforward relationship between major party competition and the number of third party candidates. As discussed in recent work by Lem and Dowling (2006), major party competition is an indicator to potential third party candidates that they can influence the outcome of an election or sway the policy of the major parties. In this regard, major party competition signals a high probability of influence or even victory to these candidates (e.g., Jesse Ventura in Minnesota in 1998). Consequently, since higher levels of major party competition, on average, result in more competitive elections, it makes sense for more third party candidates to challenge for the governorship when the level of party competition is high. Therefore, we expect that as party competition increases, the number of third party candidates contesting any election will also increase. The relationship between major party competition and major party candidate entry is more complex. In particular, we argue that the heterogeneity of the major parties should be considered. First, low levels of major party competition (i.e., when one of the two major parties is dominant) are not a guarantee of only one major party candidate because the bulk of opposition and competition occurs in the primaries (e.g., Canon 1978 on factionalism in the South). In these cases, we expect fewer candidacies in the electorally weaker major party since there is a low probability that one who declared candidacy within this party would win the general election. However, we expect several candidacies within the dominant major party since the likelihood of winning office after securing the party primary is high. The number of candidacies in the major party will depend on the level of partisan heterogeneity (i.e., the amount of policy diversity within a party); the more heterogeneous the party, the greater the number of candidates that emerge. Similarly, while higher levels of major party competition will almost certainly result in at least two major party candidates, one from each party, we do not expect the count to necessarily continue to increase. Diverse parties will likely produce more candidates within each party, so we expect the number of candidates to increase only under highly heterogeneous parties. Therefore, ideally the impact of state political competition on the number of major party candidates that contest an election is measured by the level of major party competition and the heterogeneity of the two major parties in each state. Presumably, a state with two heterogeneous parties will have more candidates, holding two-party competition constant, than a state with one (or no) heterogeneous party because more diversity in opinion results in more candidate participation.



spring 2009  /  state politics and policy quarterly   7

Social Capital. Another aspect of the social and political climate of a state is its “social capital.” In Bowling Alone, Putnam conceptualizes social capital as “connections among individuals—social networks and the norms of reciprocity and trustworthiness that arise from them” (2000, 19). These “connections among individuals,” which Putnam argues have been deteriorating for several decades now, are crucially important for political participation because most forms of political participation are group (i.e., “social”) endeavors. While the decision to run for office is not inexorably a group decision, running for office does require the aid of others. States with high levels of social capital, therefore, may encourage candidate entry by decreasing the perceived costs of entry or increasing the perceived benefit of attaining office. More specifically, norms of reciprocity and trustworthiness can promote the notion that the costs of campaigning are shared throughout the collective (i.e., an encompassing interest). Similarly, political awareness and efficacy could highlight the benefits of obtaining office, thereby increasing its perceived value. Thus, those states that foster more social capital, on average, should have a greater number of both major and third party candidates entering the electoral fray because more social connections translates into greater civic engagement and political participation generally, and candidate entry specifically.7

Election-Specific Factors and Candidate Entry State electoral institutions and aspects of a state’s social and political climate are not the only factors likely to affect candidate entry. Election-specific factors, such as the status of the gubernatorial seat, might also affect the number of individuals who choose to become candidates in gubernatorial elections by influencing the probability (P) of electoral success. The Status of the Gubernatorial Seat. The advantage incumbents enjoy when seeking reelection is well-documented (Cox and Katz 2002; Erikson 1971; Fenno 1978; Fiorina 1977; Jacobson 1987, 1990; Mayhew 1974). However, not all incumbents are the same and not every incumbent enjoys the same amount of incumbency advantage. In particular, the strength or weakness, or vulnerability, of an incumbent has a substantial effect on his or her reelection prospects. This vulnerability should have an effect on candidate participation in an election because vulnerable governors indicate a greater probability of electoral success to would-be challengers. But while we expect that a strong incumbent is likely to deter potential challengers from entering the electoral fray, it is important to distinguish between three other scenarios: (1) the presence of a weak incumbent; (2) an

8   dowling and lem

open seat resulting from an incumbent choosing not to seek reelection (e.g., retirement, death, running for president, etc.); and (3) an open seat resulting from the incumbent being legally prohibited from seeking reelection (i.e., term limits). With respect to a weak incumbent governor, we expect, on average, there to be more major party candidates contesting the election than when a strong incumbent seeks reelection, given the other major party will run a candidate and the incumbent will be likely to encounter opposition in his or her primary. An important caveat to note, however, is that while it is plausible to assume that the same logic would hold true for third party candidates, it is not clear that they should be strategic in the same manner about where and when they run, given that third party candidates have a low probability of winning an election compared to the two major parties to begin with. That is to say, the difference between a strong and weak incumbent probably does not mean as much to a third party challenger as it does to a major party challenger. As a result, we do not expect strong incumbents to necessarily limit the field of third party candidates. A clear determination of what the difference in open seat types entails for candidate entry is difficult to ascertain. On the one hand, in the case of term limits, prospective candidates know well in advance that the seat will be “up for grabs.” Given this knowledge, it is plausible that more candidates enter the electoral contest when an incumbent is term-limited out of office because of more time to prepare a campaign. On the other hand, term limits can result in lower numbers of candidates than when an incumbent chooses not to run because there is less certainty concerning the vulnerability of the incumbent party. That is, choosing not to run gives prospective candidates more information about the status of the gubernatorial seat in terms of why it is open, whereas term limits only inform prospective candidates when the seat will be open. Given its intricate nature, we attempt to more fully ascertain the role the status of the gubernatorial seat plays when it comes to candidate entry empirically.

data In order to test the proposed effects of the systemic factors outlined above on candidate entry in gubernatorial elections, we made use of a variety of data sources. Our dependent variable—the number of candidates participating in a gubernatorial election—is coded from Beyle and Jensen’s Gubernatorial Campaign Expenditures Database (2005), which contains data on all gubernatorial candidates from 1977 to 2006 (see also Jensen and Beyle 2003).8 To



spring 2009  /  state politics and policy quarterly   9

generate the dependent variables—the total number (or count) of major and third party candidates at any point in time during a state gubernatorial election-year—we simply counted each individual that is included in the dataset, every individual that spent money on a campaign.9 While the decision to become a candidate is ultimately an individual’s decision, an individual-level study similar to those done for House elections10 is beyond the scope of this article and is not feasible given our retrospective time frame (1980–2005). We contend, however, that the count of candidates provides a nice way of aggregating and analyzing individual decisions. Further, by analyzing the count, we can build a model that accounts for our systemic factors and look at how they draw out or suppress major and third party candidates. Consequently, by analyzing the number of major and third party candidates, we can get a better handle on what conditions account for there being any candidate as well as the number of such candidates when there is at least one.11

Independent Variables Table 2 recapitulates our independent variables of interest with respect to the three electoral characteristics outlined above and how each relates to Black’s (1972) decision-making calculus of potential candidates. The measurement of each of these concepts is discussed below. Electoral Institutions. As previously noted, ballot access restrictions come in many forms. This article utilizes the number of signatures required by electoral law for major and third party candidates to get on the ballot and the filing fee requirement (in dollars) for major party candidates.12 The signature requirement variable was tabulated by first calculating the actual number of signatures required by electoral law using the election code references and election results as reported by The Book of the States (various years). To make Table 2.  Independent Variables of Interest for Explaining Candidate Entry in Gubernatorial Elections, 1980–2005

P

B

C

Electoral Institutions Preprimary Party Ballot Access   Endorsement States (–)   Requirements (–) Social and Political Two-Party Competition (+) Social Capital (+) Social Capital (+)   Climate of a State Party Heterogeneity (+) Election-Specific Factors Open Seats from Term Limits (?) Presence of Incumbent (–) Incumbent Strength (–) Note: Hypothesized direction of coefficient in parentheses.

10   dowling and lem

this variable comparable across states, the number of signatures was divided by the total population. This new measure reflects the percentage of people needed to sign the petition in order to be placed on the ballot, thereby making the measure comparable across states as it controls for state size. The various legal requirements and election code references were obtained from Richard Winger (personal correspondence), editor of Ballot Access News13 and The Encyclopedia of Third Parties in America (Winger 2000, 72–95). As discussed above, we expect that as the signature and filing fee requirements increase, the total number of (major and third party) candidates will decrease because the process of becoming a candidate becomes increasingly difficult. Our other electoral law of interest—preprimary party endorsement states—is measured by a dichotomous variable. States that allow for preprimary party endorsements are scored 1 and states that do not are 0.14 Accordingly, a negative coefficient is expected. The Social and Political Climate of a State. Existing measures of electoral competition, such as the Ranney index or Holbrook and Van Dunk’s (1993) measure of competition, might be useful for gathering a basic sense of different levels of competition from state to state, but we consider them less appealing for our purposes. In particular, these indexes are based upon state legislative elections, whereas gubernatorial elections have different patterns of competition.15 Therefore, we operationalize electoral competition as the margin of victory of the winning candidate in each election. The margin of victory is calculated as the difference between the winning candidate’s vote percentage and the second-place candidate’s vote percentage. It is preferable over potential alternative measures—such as the winning candidate’s vote percentage or the difference between the winning candidate’s vote percentages and 50 percent—because it directly captures the closeness of the race between the two top candidates and is applicable to candidates winning with both a plurality and a majority of the vote. We should note, however, that measuring competition as the margin of victory could be vulnerable to election-specific idiosyncratic factors. Therefore, to approximate the overall level of major party competition in a state, we construct a three-election moving average of the margin of victory centered on the current election. To measure the heterogeneity of the two major parties, we use Poole and Rosenthal’s (1997) DW-NOMINATE scores. These scores are calculated using U.S. House legislators’ roll-call votes to calculate each legislator’s ideal position in policy-space. To calculate party heterogeneity, the amount of diversity in the state’s two major parties, we use the standard error of the



spring 2009  /  state politics and policy quarterly   11

first dimension coordinate (labeled “se_1” in the DW-Nominate data sets). Specifically, we calculate the average of these standard errors for all the House members in a state for each session of Congress. We then match these scores with the corresponding state gubernatorial election years. This calculation gives us an estimate of the average variation in a House delegation’s voting behavior (for states with only one House member, the average variation is simply that one member’s variation) for each state election year, which we use to proxy party heterogeneity.16 Social capital is measured with the index Putnam (2000, 290–3) devised for the states based upon 14 indicators of social capital.17 It ranges from low to high social capital, so we expect the higher a state is on the index the greater the number of major and third party candidates will be in an election. Election-Specific Factors. To test for the effects of incumbency and open seat types, we begin by coding two dichotomous indicators: one for incumbency (coded 1 for the presence of an incumbent, 0 otherwise) and one for open seats via term limits (coded 1 if the seat was open because the incumbent was legally prohibited from seeking reelection, 0 otherwise). As a result, the reference (omitted) category is an open seat resulting from any reason other than term limits. We also include a measure to capture the dissatisfaction of the electorate with the incumbent government because constituents often hold governors accountable for the success or failure of their state’s internal affairs (Carsey and Wright 1998; Svoboda 1995; Atkeson and Partin 1995; Niemi, Stanley, and Vogel 1995); the more vulnerable a governor, the more challengers he or she will face. Therefore, we utilize the incumbent’s previous vote share (of the total vote) to gauge the vulnerability of the incumbent governor. We then include the interaction of the incumbent’s previous vote share and whether he or she ran for reelection. By including these variables in the model, we are able to assess the effects of incumbency and different types of open seats. When the dichotomous indicators for open seats resulting from term limits and incumbency (and consequently incumbency’s interaction term with the incumbent’s previous vote share) are zero, the intercept indicates the expected number of candidates from open seats via the incumbent not seeking reelection (even though they are not legally prohibited from doing so). The coefficient on open seats via term limits reflects the change in the expected number of candidacies vis-à-vis open seats via the incumbent not seeking reelection. Similarly, the coefficient on incumbency reflects the change in the expected number of candidates when an incumbent runs in the election. The coefficient on incumbent’s previous vote share captures the effect of incumbent strength when she does not seek reelection. Last,

12   dowling and lem

the coefficient on the interaction term indicates the degree to which the strength of an incumbent, when she does run, deters or encourages wouldbe challengers. Controls. While our argument focuses on several factors influencing an individual’s decision-making calculus, there are two important caveats to note, given that our dependent variable is aggregated at the state level. First, other idiosyncratic factors—such as age, education, and career status—are likely to play an important role in whether one contests elected office in a specific election (i.e., these factors will influence P, B, or C). These idiosyncrasies, however, are automatically omitted since they cannot be included in an aggregate-level model. Although we recognize their importance in the decision-making calculus, we do not believe their omission will significantly affect the results of our model. Theoretically, these characteristics are likely to exhibit the largest effect between those who declared candidacy and those who did not. Because our analysis focuses on only the candidates that did emerge, we suspect the effect of these variables would be marginal. Additionally, there is no a priori reason to expect that the state-level variables we identify above would be strongly correlated with idiosyncratic factors such as to produce biased estimates. Second, because our analysis focuses on the count of the number of candidates in a state, we must include several state-level control variables that might influence the dependent variable but may or may not be directly tied to an individual’s decision-making calculus per se. For example, we should attempt to control for the pool of potential gubernatorial candidates in a state. We include two such measures as proxies for the pool of potential candidates: the number of statewide elected executive branch offices (other than the governor) and the size of the state legislature (lower and upper house). Both of these variables are coded from the Council of State Governments’ The Book of the States (various years). Presumably, as each of these variables increase, we should observe a greater number of both major and third party candidates. Other state-level factors that we control for include state population size and population density. Both of these variables are available from the State Politics and Policy Quarterly Data Resource18 and were updated with data from the U.S. Census Bureau’s Statistical Abstracts. Population size is used to control for the possibility that more populous states might be more attractive to candidates for reasons of prestige or larger budgets and, therefore, result in a greater number of major and third party candidates. Whereas, population density controls for the possibility that less densely populated states may have lower costs of campaigning (e.g., potentially, fewer media



spring 2009  /  state politics and policy quarterly   13

markets) and, therefore, result in a greater number of major and third party candidates. We also control for the salary of the governors, which is coded from The Book of the States (various years) and serves as a proxy for the value of the office. Additionally, for third party candidacies, we include the vote share for third party candidates from the previous presidential election to control for a general propensity of the voters of a state to support third party candidates. Last, we control for the ideology, the general “liberalness” of the citizens (e.g., Berry, et al. 1998; Wright, Erikson, and McIver 1985), of a state. We use Berry et al.’s (1998) measure of a state’s citizen ideology19 as their measure is annual (i.e., dynamic), as opposed to other static measures (e.g., Wright, Erikson, and McIver 1985), and is observed for the entire period of our study (1980–2005). If there is any relationship between a state’s ideology and candidate entry, we expect it to be positive as the citizen ideology variable runs from conservative to liberal (0 to 100); presumably liberal states are more congenial to more candidates.

results and discussion The results from negative binomial regressions estimating candidate entry into gubernatorial elections from 1980 to 2005 for major party candidates and third party candidates are reported in Table 3. Negative binomial regression is specifically designed to model count processes and, additionally, relaxes the mean-variance equality assumption that the Poisson distribution places on the data. That is, while the Poisson assumes that all events within the period are independent, the negative binomial allows for positive contagion, or the ability for the occurrence of one event (i.e., one candidate entering the electoral fray) to increase the probability of another event (i.e., another candidate entering within the same time period) (Long 1997). The negative binomial reduces to the Poisson if the mean-variance equality assumption is met (i.e., if Alpha is found not to be statistically different from zero). We begin by noting that in both equations alpha is significant. This suggests that when one major (third) party candidate enters an election we are more likely to observe another major (third) party candidate entering the same election, in the same time period, for reasons not explained by the model.

Major Party Candidate Entry With respect to the major party candidate equation, we generally find support for the electoral laws and election-specific hypotheses. Signature require-

14   dowling and lem

ments and the preprimary party endorsement system both tend to limit the field of major party candidates, but filing fees do not appear to deter these candidates from entering the electoral fray. Although there does not appear to be a significant difference in open seat types (i.e., open seats via term limits vs. the incumbent not seeking reelection), the presence of a weak incumbent encourages more major party candidates to contest the office as indicated by the coefficient on the interaction term. Thus, as the strength of the incumbent increases and the incumbent runs for reelection, the number of candidacies decreases. While the electoral laws and status of the gubernatorial seat have an effect on the number of major party candidacies, the social and political climate does not. Two-party competition and party heterogeneity are both statistically indistinguishable from zero, and while social capital attains significance, it is not in the expected direction. Furthermore, none of the control variables have a significant impact on major party gubernatorial contestation, except for the size of the state legislature, and this is not in the direction one would expect.

Third Party Candidate Entry Turning to third party candidate entry, the signature requirement variable is, as expected, significant and negative, which suggests that signature requirements are significant barriers to entry for these candidates in gubernatorial elections. The status of the gubernatorial seat, however, does not appear to be important. This finding could suggest that incumbency does not affect third party candidacies as much as major party candidacies since third party candidates are likely less concerned with winning the election. While social capital and party heterogeneity do not have an impact on third party candidacies, two-party competition is correctly signed and statistically significant.20 This finding is not surprising given that high two-party competition increases electoral uncertainty and gives third party candidates a chance to significantly alter the direction of the election or even capture elected office. Last, most of the control variables are significant in the third party model. Comparing the two models of candidate entry, we observe some interesting differences between major and third party candidacies. First, the twoparty competition variable was significant in the third party model but not in the major party model. We interpret this finding to be consistent with recent work (Lem and Dowling 2006) that demonstrates third party candidates are motivated to contest elections for reasons other than solely the chance to win the election, such as policy concerns or gaining recognition for their

Table 3.  Explaining Candidate Entry in Gubernatorial Elections, 1980–2005 Independent Variables

Major Party Candidates

Electoral Laws Signature requirement (% population) Filing fee requirement (in hundred dollars) Party endorsement states Social and Political Climate Social capital Two-party competition Party heterogeneity Election-Specific Open seat via term limit Presence of incumbent Incumbent strength Presence of incumbent × Incumbent strength Pool of Potential Candidates Size of state legislature Number of offices elected statewide (0.0114) Controls Citizen ideology Population (in millions) Population density Governor salary (in thousand dollars) Previous third party vote Constant Alpha N Wald statistic, χ2 Pseudo log-likelihood

Third Party Candidates

-0.0031 (0.0006)*** 0.0004 (0.0025) –0.4131 (0.0813)***

–0.0008 (0.0002)*** —

–0.1162 (0.0436)** 0.0003 (0.0022) 0.1365 (1.9613)

–0.0024 (0.0932) 0.0115 (0.0047)** –0.8034 (3.8753)

–0.0613 (0.0825) 0.1485 (0.3299) 0.7234 (0.4574) –0.9550 (0.5834)*

–0.2512 (0.1659) 0.1105 (0.8026) –1.0203 (1.0290) –0.7294 (1.4417)

–0.0015 (0.0004)*** –0.0001 (0.0265)*

–0.0010 (0.0007) –0.0592

0.0025 (0.0024) 0.0076 (0.0053) –0.0001 (0.0002) –0.0008 (0.0012) — (0.6958)** 1.9203 (0.2664)*** 0.0309 (0.0142) 298 192.19 –699.26

0.0071 (0.0046) 0.0087 (0.0072) -0.0008 (0.0003)** 0.0104 (0.0020)*** 1.8532

*p≤0.05; **p≤0.01; ***p≤0.001; all significant tests are one-tailed Note: Coefficients reported with robust standard errors in parentheses.

–0.0970 (0.1392)

0.5704 (0.6737) 0.2078 (0.0580) 298 118.20 –496.53

16   dowling and lem

party or themselves. Under low two-party competition, it is unlikely that a third party challenger would win the election or influence policy especially vis-à-vis a major party challenger. Conversely, when major party competition is high, third party candidates may be able to barter potential votes (i.e., the ones they would steal from the major parties) or draw more attention to their issues. Second, the strength of the incumbent is negative and significant for major party candidates, as expected, but is insignificant for third party candidates. The inference to be drawn then is that where major party candidates shy away from strong gubernatorial incumbents running for reelection, their third party counterparts do not necessarily act in the same fashion, which is probably due, at least in part, to differences in motivation for contesting elected office. While the direction and statistical significance of coefficients from the negative binomial regression are directly interpretable, gauging the substantive effects of our key variables is not straightforward. First, unlike OLS, negative binomial coefficients are not independently additive, meaning that a coefficient cannot be interpreted as the change in Y given a one unit change in X. Rather, the coefficients are inherently interactive, meaning their substantive effects may only be interpreted when considering estimates in the entire model. Second, the substantive effects of variables are not constant across their range. For example, the difference in the expected number of third party candidates when signature requirements shift from .01 to .02 percent of the population, ceteris paribus, is .29, while the difference between .02 and .03 percent is .19. To deal with these issues, we provide substantive interpretations of the results concerning incumbent strength and signature requirements using CLARIFY (King, Tomz, and Wittenberg 2000).21 In general, CLARIFY derives quantities of interest—in our case, the expected number of candidates— using simulations to derive the model’s key parameters. Through this process, the program allows us to generate the expected number of major and third party candidates at a priori defined levels of key independent variables while holding all of the other variables in the model constant at their means or modes. We can then combine the expected number of candidates over the range of a variable of interest in a single graph, which helps us discern the impact our variable has on the dependent variable. Therefore, the figures show the expected values (with 95 percent confidence intervals) given the values of each independent variable holding all else constant at their means (continuous variables) or modes (discrete variables).



spring 2009  /  state politics and policy quarterly   17

Figure 1 shows the relationship between the strength of the incumbent when an incumbent runs for reelection, with all other variables at their means or modes, and the expected number of major and third party candidates with 95 percent confidence intervals. The figure confirms what the direction of the coefficients told us: the impact of incumbent strength is different for major and third party candidacies. More specifically, major party candidates decrease as we move from a weak incumbent to a strong incumbent, from about seven major party candidates with a very weak incumbent to around four major party candidates with a very strong incumbent. However, the count of third party candidacies stays about the same (at two) as we move from a weak to a strong incumbent. Finally, there is one important similarity between the two models: the signature requirement variable performs as expected for both major and third party candidates. Again, we used CLARIFY to generate graphs of the expected number of candidates and their 95 percent confidence intervals, but this time across the range of the signature requirement variable. As shown in Figure 2, moving across the full range (0–0.02 percent) for major party candidates results in a significant decrease, about eight candidates to four or five candidates, on average (see Figure 2a). Moving across the full range (0–0.2 percent) of signature requirements also decreases the expected number of third party candidates from about 2.5 to 0.5, suggesting these requirements do limit third party candidacies (see Figure 2b). Thus, we see the effectiveness of signature requirements in limiting the field of candidates in gubernatorial elections. In other words, ballot access requirements, at least signature requirements, do restrain the amount of candidate participation and choice offered to voters in (gubernatorial) elections.

Figure 1. Expected Number of Candidates by Incumbent Strength

18   dowling and lem

Figure 2A. Expected Number of Major Party Candidates by Signature Requirement

Figure 2B. Expected Number of Third Party Candidates by Signature Requirement

conclusion The results presented in this article offer important insights into how systemic factors affect candidate entry in gubernatorial elections. By providing a general framework for studying the quantity of candidates, our model helps shed light on how state-level forces (electoral laws and the social and political climate) and election-specific forces structure individuals’ choices to pursue elected office. The results illustrate the importance of ballot access restrictions and the incumbency advantage in determining who enters the electoral arena. It is essential to point out, however, that the effect of these factors varies for individuals who choose to run under a major party label from those that campaign under minor party banners or as independents. While ballot access restrictions tend to constrain the number of both types of challengers, incumbency has significantly different effects for each type of candidate.



spring 2009  /  state politics and policy quarterly   19

Would-be third party challengers tend not to be concerned with the status of the gubernatorial seat or the strength of the incumbent holding it. On the other hand, strong incumbents seeking reelection deter major party challengers; weak incumbents encourage more contestation on the part of major party candidates. We conclude by noting that the factors that influence candidate entry, the sheer quantity of candidates—i.e., the number of choices available in an election—especially, for major and third party candidates merits further theoretical and empirical consideration. Work such as this will allow us to better understand why some elections offer voters more choices than others.

endnotes A previous version of this article was presented at the 2006 annual meeting of the Southern Political Science Association. We thank Jennifer Jensen, David H. Clark, and three anonymous referees for helpful comments on earlier drafts. We also thank Robin E. Best for generously sharing data with us. We are responsible for any errors that remain. 1. In this article, we choose to focus on gubernatorial elections for two related reasons. First, most cross-sectional office-specific differences (e.g., salary and professionalism of the office), which several scholars (see McCormick and Tollison 1978; Fiorina 1994; Hibbing 1999) often point to as increasing the propensity of individuals running for a given office, tend to be relatively small when comparing governorships. Therefore, the number of candidates in each state may be more directly comparable prima facie. Second, given the homogeneous nature of governorships, one could expect to see a lower amount of variation in the number of candidates seeking office across the states. We do, however, control for gubernatorial salary and find that it does contribute to the number of third party candidacies. 2. This figure does not include the 2003 California gubernatorial recall election. 3. Previous research concerning ballot access requirements has found them to be effective at limiting the field of candidates. For example, Ansolabehere and Gerber (1996) examine the effects of filing fees and petition requirements on congressional elections and find that more stringent ballot access requirements increase the frequency of uncontested seats and decrease the frequency of retirements. Additionally, Stratmann (2005) analyzes state legislative elections and finds that, in general, higher ballot access requirements decrease the number of both major and third party candidates. 4. We should note that most of these requirements are relatively constant over the time period under study in this article, especially for major party candidates. For a discussion of changes in third party ballot access requirements since the 1970s, see Winger (2000). 5. There is no prima facie reason the presence of preprimary party endorsements should affect the decision-making calculus of potential third party candidate entrants. 6. Depicting the amount of two-party competition in a state as an aspect of the state’s social and political climate might strike some as odd. We believe, however, that the amount

20   dowling and lem of two-party competition can straightforwardly be thought of in terms of an orientation toward political action, i.e., as a part of a state’s social and political climate. As an illustration, where one party dominates a state, the view toward politics of the citizens of that state is no doubt different from one in which two parties struggle for superiority. 7. In a sense, then, this is a direct extension of Putnam’s work in that we wish to see if social capital is related to candidate entry, a form of political participation Putnam (2000) does not consider. 8. The Beyle and Jensen data can be found at www.unc.edu/~beyle/guber.html. 9. Please note that we are not concerned with only those candidates that appear on the general election ballot. Rather, we seek to explain candidate participation more broadly. In other words, we seek to explain the total number of individuals that are candidates, for both major and third parties, at some point in time during a given election-year. 10. Most notably, the Candidate Emergence Study (CES) surveys “politically informed individuals” in order to identify the pool of “potential candidates” in recent House elections (see Maestas, et al. 2006; Maisel and Stone 1997; Stone and Maisel 2003; Stone, Maisel, and Maestas 2004). 11. Analyzing the total number of major and third party candidates does, however, suffer from some limitations. Namely, individual idiosyncratic factors that may be important for major and third party participation (e.g., personal wealth, age, etc.) cannot be accounted for. We discuss the ramifications of such omissions below. 12. In some states, the parties set the filing fee. Consequently, the fees may be different for the two parties. When this occurs we use the less stringent requirement in the model, so that potential candidates only have to jump over the lowest bar. 13. Ballot Access News can be found at www.ballot-access.org/. 14. Seven states have legal preprimary party endorsements: Colorado, Connecticut, North Dakota, New York, New Mexico, Rhode Island, and Utah (Jewell and Morehouse 1996). 15. We thank an anonymous referee for pushing us on this point. 16. We thank an anonymous referee for suggesting the use of NOMINATE scores to measure party heterogeneity. 17. The data can be found on Putnam’s website at www.bowlingalone.com/data .php3. 18. The SPPQ Data Resource can be found at www.unl.edu/SPPQ/sppq.html. 19. Their data can be found at the Inter-University Consortium for Political and Social Research (ICPSR), www.icpsr.umich.edu/; study no. 1208. For a complete description of their measurement see the article and/or the accompanying appendices available thru ICPSR. The data is also available at www.uky.edu/%7Erford/Home_files/page0005 .htm. 20. One reason for the comparatively poor performance of the social and political climate indicators in both models could be that these underlying norms have an indirect effect on the number of candidates participating in elections through some of the other variables in the model, such as the ballot access requirements or incumbency indicators. Modeling this multi-stage process is one potential fruitful avenue for future research. 21. Michael Tomz, Jason Wittenberg, and Gary King, 2003. CLARIFY: Software for Interpreting and Presenting Statistical Results. Version 2.1. Stanford University, University of Wisconsin, and Harvard University. January 5. Available at http://gking.harvard.edu/.



spring 2009  /  state politics and policy quarterly   21

references Ansolabehere, Stephen, and Alan Gerber. 1996. “The Effects of Filing Fees and Petition Requirements on U. S. House Elections.” Legislative Studies Quarterly 21:249–64. Atkeson, Lonna Rae, and Randall W. Partin. 1995. “Economic and Referendum Voting: A Comparison of Gubernatorial and Senatorial Elections.” American Political Science Review 89:99–107. Bardwell, Kedron. 2002. “Money and Challenger Emergence in Gubernatorial Primaries.” Political Research Quarterly 55:653–68. Berry, William D., Evan J. Ringquist, Richard C. Fording, and Russell L. Hanson. 1998. “Measuring Citizen and Government Ideology in the American States, 1960–93.” American Journal of Political Science 42:327–48. Beyle, Thad, and Jennifer M. Jensen. 2005. “Gubernatorial Campaign Expenditures Database.” www.unc.edu/~beyle/guber.html (December 20, 2007). Black, Gordon S. 1972. “A Theory of Political Ambition: Career Choices and the Role of Structural Incentives.” American Political Science Review 66:144–59. Canon, Bradley C. 1978. “Factionalism in the South: A Test of Theory and a Revisitation of V. O. Key.” American Journal of Political Science 22:833–48. Carsey, Thomas M., and Gerald C. Wright. 1998. “State and National Factors in Gubernatorial and Senatorial Elections.” American Journal of Political Science 42:994–1002. Cox, Gary W. 1997. Making Votes Count: Strategic Coordination in the World’s Electoral Systems. Cambridge, UK: Cambridge University Press. Cox, Gary W., and Jonathan N. Katz. 2002. Elbridge Gerry’s Salamander: The Electoral Consequences of the Reapportionment Revolution. Cambridge, UK: Cambridge University Press. Duverger, Maurice. 1951. Political Parties. New York, NY: John Wiley. Erikson, Robert S. 1971. “The Advantage of Incumbency in Congressional Elections.” Polity 3:395–405. Fenno, Richard F. 1978. Home Style: House Members in their Districts. Boston, MA: Little, Brown. Fiorina, Morris P. 1977. Congress: Keystone of the Washington Establishment. New Haven, CT: Yale University Press. Fiorina, Morris P. 1994. “Divided Government in the American States: A Byproduct of Legislative Professionalism?” American Political Science Review 88:304–16. Hibbing, John R. 1999. “Legislative Careers: Why and How We Should Study Them.” Legislative Studies Quarterly 24:149–71. Holbrook, Thomas M., and Emily Van Dunk. 1993. “Electoral Competition in the American States.” American Political Science Review 87:955–62. Jacobson, Gary C. 1987. “The Marginals Never Vanished: Incumbency and Competition in Elections to the U.S. House of Representatives, 1952–82.” American Journal of Political Science 31:126–41. Jacobson, Gary C. 1990. “The Effects of Campaign Spending in House Elections: New Evidence for Old Arguments.” American Journal of Political Science 34:334–62. Jacobson, Gary C., and Samuel Kernell. 1981. Strategy and Choice in Congressional Elections. New Haven, CT: Yale University Press. Jensen, Jennifer M., and Thad Beyle. 2003. “Of Footnotes, Missing Data, and Lessons for

22   dowling and lem 50-State Data Collection: The Gubernatorial Campaign Finance Project, 1977–2001.” State Politics and Policy Quarterly 3:203–14. Jewell, Malcolm E., and Sarah M. Morehouse. 1996. “What are Party Endorsements Worth? A Study of Preprimary Gubernatorial Endorsements.” American Politics Quarterly 24:338–62. King, Gary, Michael Tomz, and Jason Wittenberg. 2000. “Making the Most of Statistical Analyses: Improving Interpretation and Presentation.” American Journal of Political Science 44:347–61. Lazarus, Jeffrey. 2008. “Incumbent Vulnerability and Challenger Entry in Statewide Elections.” American Politics Research 36:108–29. Lem, Steve B., and Conor M. Dowling. 2006. “Picking Their Spots: Minor Party Candidates in Gubernatorial Elections.” Political Research Quarterly 59:471–80. Long, J. Scott. 1997. Regression Models for Categorical and Limited Dependent Variables: Sage Publications. Maestas, Cherie D., Sarah Fulton, L. Sandy Maisel, and Walter J. Stone. 2006. “When to Risk It? Institutions, Ambitions, and the Decision to Run for the U.S. House.” American Political Science Review 100:195–208. Maisel, L. Sandy, and Walter J. Stone. 1997. “Determinants of Candidate Emergence in U. S. House Elections: An Exploratory Study.” Legislative Studies Quarterly 22:79–96. Mayhew, David R. 1974. “Congressional Elections: The Case of the Vanishing Marginals.” Polity 6:295–317. McCormick, Robert E., and Robert D. Tollison. 1978. “Legislatures as Unions.” The Journal of Political Economy 86:63–78. Niemi, Richard G., Harold W. Stanley, and Ronald J. Vogel. 1995. “State Economies and State Taxes: Do Voters Hold Governors Accountable?” American Journal of Political Science 39:936–57. Poole, Keith T., and Howard Rosenthal. 1997. Congress: A Political-Economic History of Roll Call Voting. Oxford, UK: Oxford University Press. Putnam, Robert D. 2000. Bowling Alone: The Collapse and Revival of American Community. New York, NY: Simon & Schuster. Squire, Peverill. 1989. “Competition and Uncontested Seats in U. S. House Elections.” Legislative Studies Quarterly 14:281–95. Squire, Peverill. 2000. “Uncontested Seats in State Legislative Elections.” Legislative Studies Quarterly 25:131–46. Stone, Walter J., and L. Sandy Maisel. 2003. “The Not-so-Simple Calculus of Winning: Potential U.S. House Candidates’ Nomination and General Election Prospects.” The Journal of Politics 65:951–77. Stone, Walter J., L. Sandy Maisel, and Cherie D. Maestas. 2004. “Quality Counts: Extending the Strategic Politician Model of Incumbent Deterrence.” American Journal of Political Science 48:479–95. Stratmann, Thomas. 2005. “Ballot Access Restrictions and Candidate Entry in Elections.” European Journal of Political Economy 21:59–71. Svoboda, Craig J. 1995. “Retrospective Voting in Gubernatorial Elections: 1982 and 1986.” Political Research Quarterly 48:135–50. The Council of State Governments. Various Years. The Book of the States. Lexington, KY: CSG.



spring 2009  /  state politics and policy quarterly   23

Winger, Richard. 2000. “History of U.S. Ballot Access Law for New and Minor Parties.” In The Encyclopedia of Third Parties in America, eds. I. Ness and J. Ciment. Armonk, NY: Sharpe Reference. Wright, Gerald C., Robert S. Erikson, and John P. McIver. 1985. “Measuring State Partisanship and Ideology with Survey Data.” The Journal of Politics 47:469–89.