Electoral Rules and Legislative Particularism: Evidence from U.S. State Legislatures

Abstract

We argue that state legislative politics is qualitatively different from national congressional politics in the extent to which it focuses on localized and geographically specific legislation salient to subconstituencies within a legislative district. Whereas congressional politics focuses on casework benefits for individual constituents, state legislative politics is more oriented to the delivery of localized benefits for groups of citizens in specific areas within a district, fostering a geographically specific group connection. A primary way to build such targeted geographical support is for members to introduce particularistic legislation designed to aid their specific targeted geographical area within the district. We argue that this is primarily a function of electoral rules. Using original sponsorship data from U.S. state houses, we demonstrate that greater district magnitude and more inclusive selection procedures such as open primaries are associated with more particularism. Our findings provide strong support for a voter-group alignment model of electoral politics distinct from the personal vote/electoral connection model that characterizes U.S. congressional politics and is more akin to patterns of geographically specific group-oriented electoral politics found in Europe and throughout the world.

INTRODUCTION

Legislators’ decisions concerning how to allocate their attention have important implications for the type of public and private goods produced by the political system. Mayhew’s (1974) work on the electoral connection suggests that re-election-seeking House members offer their constituents particularized benefits for which they can claim credit. ¹ A large body of congressional research has found evidence of the electoral connection (Cain, Ferejohn, and Fiorina 1987; Fenno 1978; Fiorina 1977; Mann 1978). In a Mayhew/Fiorina world, the personal vote is a vote a relatively unsophisticated voter casts to reward a legislator’s efforts in providing individual service to that voter (or having a reputation for providing such services). This classic theory leaves little room for organized or geographically specific politics within districts. Within this model of methodological individualism, rules that limit the ability of voters to attribute individualized service delivery to a particular legislator, such as multimember districts (MMDs), should undermine legislative particularism.

When examining U.S. state legislatures, we focus on a different kind of legislative politics. We argue that because state legislative politics is qualitatively different from national politics in that it focuses on localized and geographically specific issues, the behavior of voters and candidates departs from the classic Mayhew/Fiorina work, especially in certain electoral settings. Within the model we propose, voters are attentive to their localized and geographically specific interests and willing to vote for candidates who address them effectively. Particularly in multimember districts (MMDs), candidates focus on “home bases,” subconstituencies that are usually geographically defined. Candidates, and in particular, copartisans, coordinate and collaborate, dividing different parts of a district among themselves to avoid costly electoral competition. Thus, in contrast to the classic framework, we see legislators not simply as individual actors, but as copartisans capable of collaboration. Most importantly, we view society as a dense set of geographically defined interest groups that are sufficiently heterogenous within legislative districts that even relatively small legislative districts can be divided up to different representatives elected within the district. Contrary to the Mayhew/Fiorina framework, which predicts that electoral rules that confuse the focus on individualized service delivery such as MMDs and open primaries should lead to less legislative particularism, our model predicts that greater district magnitude and more open primary procedures are associated with more legislative particularism. Sponsoring targeted and local legislation relevant to the groups in their geographical domain allows candidates in MMDs to reach the vote-share threshold for victory without incurring the costs of direct electoral competition with each other. The need to reach out to more voters in more inclusive electorates motivates legislators in states with more open primary procedures to cater to local interest groups.

This work contributes to the efforts to create a general theory of legislative elections and policy-making. We highlight how U.S. state legislatures provide a variety of electoral rules that are sufficiently similar to those in European and other settings to allow testing of “universalistic” arguments, in the process providing insights about American legislative politics that we might miss if we focused solely on national congressional politics. Scholars examining the relationship between electoral rules and legislative behavior in the U.S. have been constrained by the lack of variation in national-level electoral rules. Previous studies have focused instead on discrete electoral reforms occurring at the national level, such as the rise of primaries and the shift to the Australian ballot. By contrast, comparativists have taken advantage of the tremendous variation in electoral rules across countries to investigate the institutional determinants of the personal vote. However, the significant heterogeneity in social cleavages, party systems, and broader patterns of politics presents challenges for comparativists seeking to disentangle the effects of electoral rules on legislative particularism.

We take advantage of the unique opportunity offered by the U.S. states to test our theory. To test our theory, we compiled an original dataset of all 31,802 bills proposed in the lower houses of 29 U.S. state legislatures during the 2003–2004 legislative session. These data were assembled over the course of the last few years with the help of six research assistants. We employ a hierarchical count model, integrating district-level and state-level characteristics with individual legislator data, in the first systematic study of the effects of electoral rules on legislative particularism in U.S. state legislatures. We identify considerable variation in the amount of particularistic legislation across states. Most importantly, we find that this variation is to a large extent a function of electoral institutions, namely, district magnitude, inclusiveness of the selectorate (the body that selects candidates), and term limits. Our results provide strong support for our theory: greater district magnitude (the number of legislators elected from a district) and more inclusive selection procedures motivate legislators to propose particularistic bills in an effort to avoid costly electoral competition and to cater to interest groups. By contrast, severing the electoral connection through term limits discourages the sponsorship of particularistic legislation.

ELECTORAL INSTITUTIONS AND THE PERSONAL VOTE

The re-election-seeking efforts of U.S. House members have been the subject of tremendous scholarly research. A number of studies have found that the personal vote, the part of a candidate’s vote total that is based on his or her personal characteristics or record (Cain, Ferejohn, and Fiorina 1987), significantly affects the probability of re-election. The personal vote can be enhanced in different ways and typically involves building a reputation for providing benefits or services to the district or constituents within the district. Most of these studies suggest that the shift to more candidate-centered procedures contributed to more credit-claiming and legislative particularism. The transposition of the idea about the link between the personal vote and legislative particularism to multiseat electoral formulas with intraparty competition around the world ² is usually presented as a natural logical extension of the classic theories developed in the U.S. context, usually relying on the assumption that candidates focus on particularistic distribution to maintain a personal base that distinguishes them from their copartisans. However, a key conclusion of the comparative literature is that legislators who face competition from copartisans in MMD systems tend to target and appeal to subconstituencies within their district, geographically defined when they exist and often centered around a legislator’s hometown. Thus, the electoral dynamics identified in the comparative context presuppose the existence of organized group politics within a district. Even though largely unacknowledged, this is a significant departure from the Mayhew/Fiorina classic framework, where legislative particularism is grounded in personal appreciation for a candidate’s service to an individualized voter, an issue we further address in the following section.

Comparativists have taken advantage of significant cross-country variation in the electoral system, district magnitude (the number of legislative seats assigned to a district), and candidate selection procedures to explore the electoral foundations of particularism. A growing body of scholarship examines the effect of the personal vote on various outcomes: corruption, economic growth, budget discipline, foreign direct investment, trade liberalization, and efficacy of education spending (Chang and Golden 2005; Crisp et al. 2004; Edwards and Thames 2007; Garland and Bigleiser 2009; Gaviria, Seddon, and Stein 2003; Golden 2003; Hallerberg and Marier 2004; Kunicova and Rose-Ackerman 2005; Nielson 2003). Comparative work on legislative particularism has provided many insights, but faces serious methodological challenges that make cross-national comparisons difficult. While electoral rules are an important determinant of legislative particularism, vote-seeking incentives are the product of a number of institutional and noninstitutional factors. Conventional explanations in comparative politics focus on sociological factors such as preexisting social cleavages and ethnic divisions (see, for example, Alesina, Baqir, and Easterly 1999 and Banerjee and Somanathan 2001), cultural aspects (Putnam 1993), type of government (Hallerberg and Marier 2004), or the type of party competition (Chibber and Noorudin 2004). The significant hetereogeneity in social cleavages, party systems, and broader patterns of politics presents challenges for comparativists seeking to disentangle the effects of electoral rules on legislative particularism.

The comparative study of U.S. state legislatures provides a unique opportunity to isolate the effects of electoral incentives on legislative particularism and to test “universalistic” arguments about legislative politics and policy-making. The states allow us to hold confounding factors such as social cleavages, party systems, and broader patterns of politics constant, yet afford us significant cross-sectional variance on key features of electoral institutions such as district magnitude, candidate-selection procedures, and term limits. Somewhat surprisingly, the previous literature has not taken advantage of this unique research design, neglecting the effects of electoral institutions on particularism. Several recent works take advantage of the proliferation of online voting records and sophisticated webscraping techniques have reduced the costs of collecting comprehensive state-level data (Clark et al. 2009; Shor et al. 2010), but those focus on the effects of legislator demographic characteristics on bill initiation and policy success (Bratton and Haynie 1999; Whitby 2002) or the role of parties in state legislatures (Battista and Richman 2011; Jenkins 2008; Wright and Schaffner 2002). Examining a sample of 165,000 bills spanning 120 years in 13 different states, Gamm and Kousser (2010) identified a considerable amount of particularistic legislation and found that particularistic legislation was related to the level of party competition in the legislature, with one party dominance associated with more particularistic bills, and members’ salaries (with higher compensation associated with more particularistic bills). While closest to our analysis, Gamm and Kousser’s study does not investigate how variation in electoral rules across states affects members’ incentives to introduce particularistic legislation, which is our primary interest.

THEORETICAL EXPECTATIONS

District Magnitude and Legislative Particularism

The comparative literature has transposed theories about the connection between the personal vote and pork barrel politics from the American context to open list PR systems around the world. A number of scholars argue that where electoral rules encourage intraparty competition, as in open list PR where candidates are seated in accordance to the individual preference votes received, the incentives to build a personal vote increase with district magnitude because the number of co-partisans from which a candidate must distinguish herself increases. Compared to SMD candidates, MMD candidates in open-list must appear distinctive among a broad field of ideologically similar candidates (Ames 1995a; Carey and Shugart 1995; Schiller 2000). Because voters cannot rely on the party label to distinguish them, co-partisans tend to focus on the distribution of particularistic benefits to build a personal base. Most empirical studies of systems with MMDs and intraparty electoral competition find support for the claim that legislative particularism increases with district magnitude. This was established in the context of Italy (Golden and Picci 2008), Colombia (Crisp and Ingall 2002), Estonia (Tavits 2010), Indonesia (Allen 2012), and Brazil (Ames 1995b, 2001; Mainwaring 1991, 1999; Samuels 2001). While this extension from the U.S. to the comparative context is seemingly natural, comparative research has largely neglected the difficulty of attributing particularistic services and outcomes to individual legislators in MMDs under personalistic electoral rules. While voters in SMDs in the U.S. context can attribute the works of staffers to “their legislator,” voters in MMDs should be naturally uncertain about precisely which legislator serves their interests or delivered a particular project. ³ In the existing literature, the motivations of the voter to choose a particular candidate are assumed or set aside; theories focus on the individual candidate or legislator.

Regarding the motivations and strategies of candidates, Cox (1990) and Myerson (1993) show that the low vote share threshold needed to guarantee a seat in high-magnitude PR systems incentivizes candidates to target small subconstituencies within the district. Appealing to a broad cross section of an MMD might make a candidate vulnerable to a challenger that targets private goods to a subconstituency of a minimal size necessary to win a seat (Myerson 1993). While theories do not necessarily imply that the targeted constituencies are geographically defined (Myerson 1993), the geographic concentration of a candidate’s electoral support is a typical pattern in MMD systems with intraparty competition. Colluding and dividing the district subconstituencies is an efficient way for candidates to solve a coordination problem and avoid “expensive” electoral competition (see Katz and Mair 1995). While politicians can and sometimes do target identities based on ethnicity, occupational group or class, targeting constituencies concentrated in a particular geographic area has the lowest campaigning costs (Ames 1995b). Studies finding evidence of geographically defined subconstituencies or bailiwicks in MMD systems, often centered around a candidate hometown, include Ames (2001) on Brazil, Hirano (2006) on Japan, and Crisp and Desposato (2004) on Colombia. ⁴

Whereas candidates in the early 20th century United States depended upon party leaders to gain access to the ballot, progressive era reforms resulted in the dominance of primary elections. Currently, most states employ different types of primary procedures and candidates must compete against copartisans in order to gain access to the general election ballot (Bibby and Holbrook 1999). ⁵ Although most legislators are elected from single-member districts (SMDs), as of 2013, 10 states continue to use MMDs to elect at least some legislators. While there has been some research examining how multimember state legislative districts influence representation, much of this work focuses on whether MMDs influence the ideological positioning of legislators through analysis of roll-call votes (Adams 1996; Niemi, Hill, and Grofman 1985; Richardson, Russell and Cooper 2004) or whether district magnitude affects the legislative success of women and minority representatives (Bratton 1999). We seek to understand how MMDs in the U.S. influence legislative particularism. While this is a largely undertheorized question, the extensive literature on the personal vote and the electoral connection in the U.S. context implies that MMDs should be associated with less legislative particularism. In the classic Mayhew/Fiorina framework, the “personal vote” is a vote an individual casts for appreciation of a representative’s efforts in providing personal services or benefits to a voter, which in turn allows incumbents to win large margins of victory beyond what would be predicted by their partisan or group affiliations. Electoral rules that reduce the ability of legislators to develop a reputation for particularistic benefits to constituents and reduce the ability of voters to precisely reward such legislative behavior should reduce legislative particularism. In MMDs (due to increased district magnitude), voters naturally may be confused about precisely which legislator is serving their needs and should be credited for particularistic services and outcomes. ⁶ Given the difficulty of attributing particularistic services and outcomes to individual legislators, one might expect that legislators in MMDs operating under plurality rules would stress less casework and personal vote strategies. Multimember districts present free-rider problems which disincentivize legislators to sponsor territorially oriented legislation (Ashworth and Bueno De Mesquita 2006, Lancaster 1986). With more than one representative being able to claim credit for a given local project, and given voters’ general lack of awareness about legislative activity, it is unlikely that voters can easily reward a legislator for delivering benefits to a particular district. The few empirical studies that investigate the relationship between district magnitude and the local responsiveness of legislators in U.S. state legislatures are based on survey study and have inconclusive results. While Jewell (1982) finds that MMDs discourage legislators from responding to constituents, Freeman and Richardson (1996) find the opposite.

Other more recent studies are at odds with the claim that legislative particularism is lower in MMDs. Snyder and Ueda (2007) argue that the free-riding incentives that may exist in MMDs are counterbalanced by other factors. MMD representatives who represent “natural economic communities” such as cities or counties are motivated to collaborate on collective projects, which allows them to wield larger power in the legislature. The pattern of resource distribution in the U.S. is consistent with this theory: they find that MMD-represented municipalities receive more state funding per capita. Similarly, Kirkland (2012) argues that in order to compensate for their limited incumbency advantage, legislators in MMDs form coalitions to maximize credit-claiming opportunities and benefits for their district, which they all share. Kirkland (2012) finds that MMD legislators sharing a constituency in North Carolina cosponsor more legislation together than others, especially, but not only, when it comes to nonpartisan issues. The direct relationship between district magnitude and legislative particularism has not been tested across the American states.

Drawing on the comparative literature, we offer a novel theoretical perspective that predicts, contrary to the electoral connection literature, that geographically targeted bills should be more frequent in MMDs in the U.S. MMD elections in the U.S. follow plurality voting rules, which, combined with the use of primaries generates intraparty competition and a dynamic similar to that in open-list proportional representation and other electoral systems where voters can indicate preference for individual candidates. We argue that similar to their counterparts in PR systems with intraparty competition around the world, candidates, and in particular, “copartisans” coordinate and divide different parts of a district among themselves, targeting small groups within the district, subconstituencies that are usually but not exclusively geographically defined and organized, and often ones in which candidates reside. In turn, instead of individual voters casting votes for personal or particularistic benefits a legislator is delivering to him/her à la Fiorina, in our model “group” voters support candidates furthering subconstituency-specific interests. Recent work in American politics (Dodd 2015) suggests that voters are capable of rational policy voting and candidates are capable of addressing such constituency-specific calculations and issues. Grimmer (2013) demonstrates that U.S. Senators strategically emphasize particularism and district/state work in contexts where there may be an ideological mismatch between the member and constituents and where electoral competition is high. Sponsoring targeted and local legislation relevant to the groups in their geographical domain allows candidates in MMDs to reach the vote-share threshold for victory without incurring the costs of direct electoral competition with one another. As the number of seats per district, i.e., district magnitude, increases, state legislators have even greater incentives to sponsor targeted and local legislation for which they could claim credit with voters in their geographical territory. This interaction presupposes the existence of meaningful parties that allow candidates to engage in collaboration and coordination and, most importantly, on a range of geographically specific groups within a district. Thus, the voter-group alignment model we propose departs from the classic Mayhew/Fiorina framework where the electoral connection between the voter and the candidate is based on personal appreciation for service to an individualized voter.

We argue that the distinct character of state legislative politics fosters a more group-oriented politics than the politics examined in the Congressional literature. A personal electoral connection à la Mayhew/Fiorina exists and is crucial in SMDs with majoritarian electoral rules, and especially in legislative arenas where casework is important and may deliver significant benefits to citizens, ⁷ as in the U.S. Congress. While the national legislature tends to focus more on broad national policy issues, state legislatures tend to focus on local affairs because American federalism gives each state complete authority over the local governments within it (Frug 1980, 1999; Gamm and Kousser 2010). ⁸ Because in state legislatures many issues are susceptible to local and geographically specific concerns, the use of MMDs and the intraparty competition and coordination it generates may activate localized group voting. Thus, we claim, the relevance of the Mayhew/Fiorina model of legislative politics depends on context.

Hypothesis 1: As the number of seats per district (i.e., district magnitude) increases, legislators have incentives to distinguish themselves from their copartisans and therefore will initiate significantly more local and targeted bills.

Inclusiveness of the Selectorate and Legislative Particularism

While the effect of primary types on legislator ideology has been the subject of significant research (see, for example, Gerber and Morton 1998 and McGee et al. 2014), its impact on the kinds of bills being sponsored by members has been neglected. Primaries in the U.S. come in several types, varying in who comprises the selectorate, the body that selects candidates. Due to variation in party enrollment procedures across states, classifying states by primary type is more complicated than the simple open versus closed distinction (Carr and Scott 1984; Finkel and Scarrow 1985; Jewell 1981). We argue that candidate selection and, in particular, its impact on legislative voting are undertheorized in the context of U.S. state legislatures and yet have significant effects.

In the world as theorized by Mayhew/Fiorina, more open primary procedures should be associated with weaker personal connections between voters and representatives and less individualized constituency service. In open primary states, electoral success depends not only on the support of a legislator’s partisan constituency; there is greater uncertainty about composition of the electoral coalition of a candidate. In this environment, legislators are disincentivized to provide individualized service delivery and should instead focus on more general policy concerns. While the personal vote/electoral connection framework predicts less particularism under more inclusive primary procedures, the voter group-alignment model we propose predicts that more inclusive primary procedures should lead to more legislative particularism. We argue that legislators in states with more open primary procedures focus more on geographically specific interests, in particular, on organized interest group interests. Candidates selected by an inclusive selectorate do not need to be loyal team players in order to be successful politicians; instead they have incentives to cater to their inclusive selectorates. This might shorten the time horizon of politicians, motivate them to act in a manner that disregards the interests of their party, and, most importantly, lead to an effort to obtain more financial resources in order to reach more voters. Broadening the selectorate could thus empower well-organized local interest groups (Rahat and Hazan 2001, 313–4). Studies of the Israeli case support the assertion that the transition to more open candidate selection procedures empowers local interest groups. According to Hofnung (2003, 67), the greater need for campaign funding under more open procedures increased the influence of money donors. Where electoral rules make individualized service/benefit delivery to constituents an inefficient strategy, delivering particularized benefits to local interest groups that benefit local subconstituencies indirectly could be an efficient way to reach out to the broad selectorate and be successful electorally. ⁹

While candidate-selection procedures are receiving growing attention, their impact on legislative voting behavior has not been examined systematically in comparative fashion using individual-level data, to our knowledge. Here we address this gap in the previous literature and test the voter-group alignment model in the U.S. context. The exact location of primaries on the exclusiveness/inclusiveness continuum depends on the specific restrictions in state laws (Kolodny and Katz 1992; Ranney 1981). At one extreme are systems where the whole electorate has the right to vote for a candidate. At the other extreme are the most exclusive selectorates where one party leader nominates the candidate. All types of American primaries are to some extent inclusive procedures located in the “electorate zone” (i.e., closer to the inclusive end of the continuum); however, there is variation in the degree of their inclusiveness. The blanket primary, where every registered voter can vote for candidates from any party, would be located toward the most inclusive end of the “electorate zone.” By contrast, closed primaries, which require voters to register their party affiliation before or on the day of the primaries, are located toward the exclusive end of the “electorate zone” (see Rahat and Hazan 2001, 301–2).

Hypothesis 2: More inclusive primary procedures provide incentives for legislators to propose more targeted and local legislation: legislators operating under open and semiopen primary systems will propose significantly more particularistic bills than legislators operating under closed or partially closed primaries.

Term Limits and Legislative Particularism

The assertion that legislative particularism varies in accordance to the electoral system and the personalistic incentives it provides rests on the assumption that legislators are in fact seeking re-election. This is generally the case in most democratic legislatures throughout the world, yet the adoption of term limits offers an instance in which the electoral connection is severed since legislators are restricted from seeking re-election. Scholars have generally agreed that term limits render legislative particularism useless to candidates, regardless of the specific character of the broader electoral system, since legislators no longer have incentives to cultivate a strong personal reputation (Carey, Niemi, and Powell 1998; Herron and Shotts 2006). Even in the case of “termed-out” legislators, they may face different incentives depending on their career paths. For instance, examining 210 legislators in the Michigan legislature from 1982 to 2000, Tothero (2003) finds that many members sought local offices with some also seeking statewide offices. Despite the paucity of research on the relationship between term limits and legislative particularism in state legislatures, there have been studies that attempt to gain leverage on this question through systematic analysis of retirees in Congress. ¹⁰ The empirical evidence is somewhat mixed. ¹¹ Taking advantage of the institutional variation in term limits restrictions in the U.S. states, we test these theories at the state level. Currently, 15 states have term limits and their staggered nature enables us to test how term limits affect individual members’ incentives to initiate particularistic legislation.

Hypothesis 3: Legislators who may no longer seek re-election due to term limits will initiate significantly fewer local and targeted bills than re-election-seeking members not subject to term limits.

Because term limits influence members’ incentives, their electoral circumstances may moderate the relationship between candidate selection methods (i.e., inclusiveness) and legislators’ decisions to sponsor local or targeted legislation. Retiring and termed-out legislators might not be motivated to cater to special interest in order to secure financial resources and the ability to appeal to a broad cross section of the electorate in the next election. However, legislators who are termed-out yet pursuing another office may be motivated by other constituencies. These other constituencies may be broader (e.g., statewide) or narrower (e.g., city), and there will be special interests associated with each of these. Thus, we expect that not all members will be uniformly influenced in the same manner by the inclusiveness of the selectorate. For instance, members for whom re-election is not a concern (i.e., termed-out and not seeking another public office and retiring members) may not be as likely to sponsor particularistic legislation in open systems (as members seeking public office) since the electoral connection has been severed. In that case, we would expect an attenuation of the relationship between the inclusiveness of the selectorate and the number of local and targeted bills sponsored for term limited members who are not seeking another office.

Hypothesis 4: The relationship between candidate selection method (i.e., inclusiveness) and the sponsorship of local or targeted legislation is conditional on members’ electoral circumstances (i.e., whether they are termed out, seeking another office, etc.).

DATA AND METHODS

To test these hypotheses, we constructed a unique dataset of state legislative electoral rules and bill initiation for 29 state houses (which vary in terms of electoral rules, party control, and region) during the 2003-2004 legislative session. We compiled the central database for this research, which focuses on the sponsorship of particularistic legislation, from state legislative websites. Legislative records for each chamber are maintained by the clerk’s office and in most states have been posted online. Since these are maintained by the clerk’s office of each chamber, there is no consistency in what records are posted and where they are posted. For our purposes, this required us to investigate each website to see how the data were structured (e.g., SQL database, FTP, etc.) and where they were maintained. To obtain all of the bill texts as well as bill sponsorship information, we create a Perl script for each state legislative website. We used the LWP package (short for “Library for WWW in Perl”) and looped through, in some instances, thousands of pages of bill text (e.g., some websites posted a bill per webpage), pulling the text, and stripping the html markup. As an example, in only 10 lines of code, we created a script that would scrape about 10, 000 Georgia House bills in approximately 3 minutes.

Although many state legislative records are now online, we concentrated on a sample since our research requires time-intensive content analysis and coding of legislation. We selected the states which maximize variation along an array of dimensions including the electoral characteristics discussed above as well as: region, legislative professionalism and careerism, state incomes, and racial/ethnic composition. ¹² We identified every bill introduced during the 2003-2004 legislative session. We collected comprehensive data related to the bill’s introduction, including the title and description of the bill, the bill sponsors, the bill text, and any other bill history data, including votes and passage. Based on the bill text, we categorized each bill as general, local, and/or targeted legislation. We used the bill text rather than summaries because summaries often excluded key information concerning the bill’s scope. We tasked 6 research assistants with coding the data. A total of 31,802 bills were coded. Research assistants first read each bill to determine whether the bill was a general or local piece of legislation. They then consulted the state legislative district map to determine whether the local bill was also “targeted,” which meant that the bill concerned the particular district of the legislator who sponsored the bill. In each case, at least two researchers coded the bills. The intercoder reliability (Cohen’s Kappa) was 0.98, which gives us confidence in our codebook and coders.

To ensure consistency in coding, we developed a codebook to guide the research assistants in coding bills. In the Online Appendix, we provide more detailed information about the codebook. The codebook draws from Gamm and Kousser’s (2010) historical analysis of legislative particularism in the state legislatures. The codebook includes careful instructions for distinguishing general bills from local bills and local bills from targeted legislation. Like Gamm and Kousser (2010), we take a broad perspective on identifying particularistic legislation, including all bills pertaining to particular local governments. We argue that local bills, regardless of whether they contain distributive benefits, bring some benefits back to the district and are introduced for the same reasons—electoral motivations. Although local bills may not be passed, they represent clear attempts to deliver something to voters or as a signal to constituents.

Local bills were relatively easy to identify because these bills refer, by name or classification (e.g., “County of San Bernardino” or “the Los Angeles Unified School District”), to a specific, geographically distinct area of the state. For example, one bill, introduced by California State Representative Sarah Reyes, sought to appropriate an unspecified amount of money for an emergency loan to the West Fresno Elementary School District (AB 38, prefiled on December 2, 2002). This legislation would first be coded as a local bill and then coders would determine whether the West Fresno Elementary School District resides in the legislative district of Rep. Reyes. Since her district does include the West Fresno Elementary School district, the bill is coded as “targeted legislation.”

Bills not deemed local were classified as general bills, which relate to “general issues” of state or local concern (e.g., taxation, health care policy, education). General legislation may concern local governments or local affairs but do not specify a particular local government. These general bills may include bills that concern government agencies, such as cities, towns, villages, municipalities, water districts, school districts, sewer districts, park districts, counties, and local courts and commissions. These also include bills that address issues that primarily affect local governments. Thus, we coded bills as “general legislation” if they addressed local bonds, property taxes, taxation, intrastate boundaries, local school policy, local government employees, local utilities, local government infrastructure, and government services typically supplied at the local level. One example of a general bill sponsored by Indiana State Reps. Kersey, Adams, and Neese would require “a city fire department or a city or county police department to give a hiring preference according to the following priority: (1) A war veteran. (2) A person whose mother or father was a full-time, fully paid firefighter or police officer who died in the line of duty” (House Bill No. 1089, introduced on 01/07/2003).

Although many political pundits, journalists, and some politicians have argued that particularism or pork is “bad” for democracy, others argue that pork is often a relatively small proportion of the budget overall and can have a profound impact upon the communities that are affected. Legislators’ decisions of how and whether to sponsor local/targeted legislation or general bills has important consequences for the kind of representation enjoyed by citizens. Indeed, some targeted bills could have benefits beyond the particular organization or entity receiving the funds. For instance, California Assemblywoman Jacqui Irwin secured $500,000 for lab equipment to create an engineering degree program at California State University Channel Islands in Camarillo. This not only benefits the university and the city but also California as investing in engineering programs should help strengthen the economy and the workforce.

We identify considerable variation in the percentage of local, targeted, and general bills sponsored. Table 1 shows the percentages of local and targeted bills initiated during the 2003-2004 session in each chamber. The percentages of particularistic legislation reveal that chambers in our sample ranged from almost 1% of bills initiated attending to local issues to a high of around 62% of bills. For targeted bills, we find a range of 0.1% and about 61%. In Georgia and Louisiana, the majority of the agenda focused on local legislation. The percentage of local and targeted legislation in Georgia, Louisiana, and Ohio is one standard deviation or more above the average for all 29 states in our sample. The percentage of both local and targeted legislation in Delaware, Georgia, Illinois, Louisiana, Maryland, Minnesota, Missouri, North Carolina, New Hampshire, New Jersey, Ohio, Texas, Vermont, and West Virginia is above the median percentage. The frequency of local legislation is highly correlated with the frequency of targeted legislation (0.98).

Table 1. Local and Targeted Bill Sponsorship Patterns in U.S. State Legislatures

Table 2 presents the key electoral rules that we examine. To investigate the effect of district magnitude on legislators’ propensity to initiate local and/or targeted legislation, we gathered data on district magnitude for each legislator. The Project Vote Smart database contains information on the legislative district that each member represents. The National Conference on State Legislatures provides information on which states have multimember districts. We cross-referenced those data to ensure that we did not mistakenly code a SMD that perhaps had a representative leave midterm who was replaced by another person as a MMD with district magnitude of 2. In most of the multimember districts, district magnitude is 2, with the exception of New Hampshire and West Virginia, where district magnitude ranges from 1 to 14 and 1 to 7 respectively. There are some states that used mixed systems consisting of some races elected through single-member districts and others using multimember districts; these include Georgia, Maryland, and Vermont. In addition to the eight states that employ multimember districts for some seats, Idaho and Washington have postdesignated MMDs (also sometimes called seat-designated MMDs). Some of the districts in Maryland and South Dakota are also postdesignated MMDs. Postdesignated MMDs involve an election in which candidates run for a particular post, or seat, within the district, so the district is subdivided into Post A and Post B, for instance, and candidates running in Post A are not in direct competition with those in Post B for the 1 seat available in their post. According to Hamm and Moncrief (1999, 148), “In effect, such systems operate as a series of single-member elections within the same district.” Moreover, Niemi, Jackman, and Winsky (1991, 97) state that “MMDs with positions can be regarded as equivalent to SMDs. . . .” In line with the consensus in the literature (see Cox 1984, Hamm and Moncrief 1999, Niemi, Jackman, and Winsky 1991) we consider the postdesignated districts not pure MMDs because they serve as de facto SMDs taking place within the same geographic district.

Table 2. Electoral Incentives (2002)

We operationalize the degree of inclusiveness or exclusiveness of the selectorate through candidate selection procedures. We created a contextual variable for the inclusiveness of the selectorate that ranges from 0 (most exclusive) to 4 (most inclusive). Although all of the states in our sample use a primary to nominate candidates, these primaries vary a great deal in terms of their inclusiveness of voters. Louisiana, which employs a “blanket primary” is at the most inclusive end of the continuum (“Inclusiveness”=4). Open primaries employed in Georgia, Idaho, Michigan, Minnesota, Missouri, Montana, North Dakota, and Vermont, are also fairly inclusive procedures (“Inclusiveness”=3). At the other end of the continuum are closed primaries used in Delaware, Kansas, New Jersey, Nevada, Ohio, Oklahoma, and Pennsylvania (“Inclusiveness”=0). Partially open and partially closed primaries employed in the rest of the states are procedures with intermediate levels of inclusiveness, with partially open primaries closer to the inclusive end of the continuum (“Inclusiveness”=2 and 1, respectively).

To examine our third set of hypotheses, we also collected data on which legislators were currently affected by state legislative term limits. During the time frame of this study, 15 states had adopted term limits. Although 11 states in our sample had adopted term limits prior to our time period, members from two states (Louisiana and Nevada) were not yet affected under the rules, which imposed term limits after 12 years of service. ¹³ Therefore, we treat Nevada and Louisiana members as not being affected by term limits since they were not yet in effect during the time period under examination. Additionally, in 2002, the Idaho state legislature repealed the statutory term limits that were put into place by citizen initiative in 1994. The National Conference on State Legislatures provided us with data on which members were affected by term limits and whether those members ran for another elective office. We merged these data with our master bill sponsorship dataset. We created a series of dummy variables for whether the member was “termed out” and not seeking any other public office (“Termed Out,” 1=“termed out” & not seeking another public office, 0=otherwise), whether the member was “termed out” and seeking another public office (“Other Office,” 1=subject to term limits and seeking another public office), whether a member was retiring on their own accord (“Retiring,” 1=retiring, 0=otherwise), and the baseline is continuing members. ¹⁴

We include a number of individual-level control variables in our model. First, we include a measure of seniority, which is simply the number of years the legislator has served in the legislature. Previous research has found that members with longer tenure in office are more likely to initiate targeted legislation (Taylor-Robinson 2006). Studies also suggest that parochialism is incentivized in rural districts, and therefore, we would expect members who represent rural districts to sponsor more particularistic legislation (Cox and Morgenstern 2001; Samuels and Snyder 2001; Suiter and O’Malley 2013). The variable rural takes on a value of 1 for legislators from rural districts, and 0 otherwise. We also include dummy variables for the legislator’s gender (Taylor-Robinson and Heath 2003) and race (Bratton and Haynie 1999). The variable female takes on a value of 1 for women legislators, and 0 for males. The variable white takes on 1 if non-Hispanic white and a 0 if nonwhite or Hispanic. Finally, we include a measure of electoral competition in the general election, which is measured by the difference in the vote percentage received by the winning candidate and the vote percentage received by the losing candidate.

It is important to hold confounding contextual factors constant. One important factor that we control for is legislative professionalism. We expect higher salaries to encourage more particularistic legislation by heightening the incentives to get re-elected (Gamm and Kousser 2010, 156). Another relevant factor is the length of legislative sessions. According to Gamm and Kousser (2010), longer legislative sessions should be associated with more statewide bills since longer sessions give legislators more time to work on more complex legislation. Longer sessions motivate legislators to develop and maintain a good reputation within the house through introducing legislation with broader impact. We expect the percentage of targeted bills to depend on whether there is strong two-party competition. Previous works suggest that stronger party competition will motivate legislators to focus on statewide policies, while single-party dominance produces incentives for parochialism (Gamm and Kousser 2010; Key 1949).

We also include a measure of Income Per Capita. Previous research has found that income has a significant impact on state public goods provision (Dye 1966; Hofferbert 1966; Winters 1976). We control for state house turnover. Over the past century, state legislatures have transitioned from high turnover to low turnover institutions (Polsby 1968). We expect this shift to affect the prevalence of particularistic legislation. Legislators returning to the capitol session after session should be more inclined to identify themselves more closely with the state and may have a broader outlook. Representatives in legislatures with low turnover rates should also be motivated to initiate bills that enhance their present and future influence in the state house, i.e., general bills versus targeted bills (Gamm and Kousser 2010, 7). We include dummy variables to control for whether the state has adopted any limitations on the sponsorship of local bills. We consulted with both the National Conference on State Legislatures and the individual state constitutions to assess specific rules restricting whether members may sponsor local legislation. ¹⁵ In instances where the state placed a limit on sponsorship of local legislation, we coded those states with a 1 and in states where no such limits exist, we coded the state as a 0. In our sample of state houses, only four chambers limit the overall number of bills a member may initiate (Indiana, Minnesota, North Dakota, New Hampshire, and West Virginia). These chambers were coded 1 with the others coded 0 and correspond to the Constitutional Limitations variable in the models. Finally, we also include a dummy variable for statutory bill limitations, which impose a limit on how many total bills a member may sponsor regardless of whether they are general or local. In the Online Appendix, we provide summary statistics for each variable.

In addition to the individual-level and contextual effects, one benefit of multilevel modeling is the ability to examine how and whether state-level contextual features moderate the effect of individual-level features on legislator’s likelihood to sponsor local or targeted legislation though the inclusion of cross-level interactions. We argued earlier that the selection process shapes members’ incentives to cultivate a personal vote. This effect may not be constant for all members, however. Where parties are weakest (open primaries), we may expect to find that members of the minority party have even greater incentive than their majority party counterparts to initiate particularistic legislation to cultivate a personal vote to overcome their inherent disadvantage in getting general policy-related measures passed in the chamber (Hasecke and Mycoff 2007). We also expect that not only would a one-party dominated system incentivize particularism but that in such a system, members of the minority party may have even more incentive than their majority party counterparts to cultivate a personal vote through the sponsorship of particularistic legislation (Key 1949). Therefore, we would expect a negative relationship between majority party status and the likelihood of initiating local or targeted bills conditional on the level of party competition. Additionally, we expect that a member’s electoral circumstance will moderate the effect of the selection system (i.e., inclusiveness) on the sponsorship of local and targeted legislation. We include the following cross-level interactions to take this into account: termed-out x inclusiveness, higher office seeking x inclusiveness, and retiring x inclusiveness.

STATISTICAL MODEL

We develop a varying-intercept, varying-slope hierarchical model with cross-level interactions, in which legislators are nested within states. The response variable is the number of local (or targeted) bills sponsored by the legislator, which takes on non-negative integer values, so specifying a count model is more appropriate than OLS (Long 1997). In count models, we set out to predict the dependent variable as a function of the data, but to obtain unbiased estimates, we must take into consideration the discrete and non-negative characteristics of the dependent variable. Thus, we specify the following model:

(1) $$\begin{equation} Y_{ij} = \beta _{0j}+\beta _{1j}X_{ij}+\cdots +\beta _{Nj}X_{ij}+\epsilon _{ij},\\ \end{equation}$$

(2) $$\begin{equation} \beta _{0j} = \gamma _{00} +\gamma _{01}Z_{j}+\mu _{0j}, \\ \end{equation}$$

(3) $$\begin{equation} \beta _{1j} = \gamma _{10} +\gamma _{11}Z_{j}+\mu _{1j}, \\ \end{equation}$$

(4) $$\begin{equation} \beta _{2j} = \gamma _{20} +\gamma _{21}Z_{j}+\mu _{2j}, \\ \end{equation}$$

(5) $$\begin{equation} \beta _{3j} = \gamma _{30} +\gamma _{31}Z_{j}+\mu _{3j}, \\ \end{equation}$$

(6) $$\begin{equation} \beta _{4j} = \gamma _{40} +\gamma _{41}Z_{j}+\mu _{4j}, \\ \end{equation}$$

(7) $$\begin{equation} \beta _{5j} = \gamma _{50} +\gamma _{51}Z_{j}+\mu _{5j}. \end{equation}$$

In the first equation, Y_ik indicates the number of local bills sponsored by legislator i from state j. The individual-level parameters are given by Σβ _t X_ij . The second equation represents the random intercept, β_0j , capturing the mean rate of cosponsoring targeted legislation for state j as a function of Z_j . Thus, if γ₀₁ is positive, the average rate of proposing targeted legislation is higher in systems with open primaries for instance. Equations (3) and (4) estimate the cross-level interactions. Equation (3) estimates the relationship between majority party status and the propensity to sponsor local or targeted legislation conditional on the candidate selection system. If γ₁₁ is positive, members in the majority are more likely to sponsor targeted or local bills under open endorsement systems (i.e., the open primary). If γ₁₁ is negative, the effect of majority party status is attenuated in open endorsement systems. Similarly, in Equation (4), if γ₂₁ is positive, then the effect of majority party status on sponsoring local or targeted bills is enhanced in one-party dominant systems. If γ₂₁ is negative, then the effect of majority party status on sponsoring local or targeted bills is diminished in one-party dominant systems.

The effect of the nomination system on members’ incentives to sponsor particularistic bills may also be conditional on members’ electoral circumstances. We capture this interactive effect in Equations (5)–(7). We expect that γ₃₃ will be negative, meaning that the effect of the candidate selection system on the incentive to sponsor particularistic legislation is significantly reduced for members termed out and not seeking another elected office because the electoral connection is severed. We also predict a negative effect for members who are retiring from office. However, for members pursuing another elective office, the predictions are less clear. Although the electoral connection with current constituents is severed, incentives to demonstrate adherence to voters through particularism may remain.

RESULTS

Individual-legislator and District-level Variables

We present the main results of the hierarchical negative binomial regression model (NBRM) in Table 3. We present the estimates of the hierarchical NBRM for targeted bills in column 2, and column 4 contains the estimates of the hierarchical NBRM for local bill initiation. To facilitate interpretation of these effects, we also present the first differences in column 3 for targeted bills and column 5 for local bills. The first differences provide a means to assess the magnitude of the effect of each variable on the outcome, sponsorship of local and targeted bills. These are estimated using CLARIFY which essentially holds all covariates constant (at a certain value of theoretical interest) and the independent variable of interest was allowed to take on the range of values that we specified (King, et al. 2000). For continuous variables, we estimated the effect of moving from the 25th percentile to the 75th percentile on the expected number of targeted (or local) bills sponsored holding other continuous variables at their mean and dummy variables at 1. For dummy variables, we estimated the effect of moving from 0 to 1 on the expected number of targeted (or local) bills sponsored when holding all other variables at their mean (or 1 in the case of dummy variables).

Investigating the individual-level covariates, we find considerable variation in legislators’ decisions to sponsor particularistic legislation. The results demonstrate that district magnitude influences legislative incentives to initiate particularistic bills. The positive coefficient indicates that as district magnitude increases, the likelihood of legislators to sponsor targeted and local bills increases. We find that in moving from a single-member district to a district magnitude of 3, we expect members to sponsor about 12 more pieces of targeted legislation and about 11 more local bills. These findings go against the classic Mayhew/Fiorina framework, according to which electoral rules that confuse the personal connection between a candidate and a voter lower the incentives for legislative particularism, but are consistent with the implications of our claim that candidates and in particular, co-partisans, divide a district into subconstituencies and cater to those small groups by sponsoring targeted legislation. More broadly, the results are consistent with the personal vote literature which finds that district magnitude intensifies personalistic incentives in systems with strong intraparty competition (see, for example, Chang 2005; Chang and Golden 2007; Edwards and Thames 2007).

We also find that term limits influence legislators’ decisions to initiate particularistic legislation. Specifically, we find that termed-out members who will not pursue another public office initiate almost six fewer local or targeted bills (than re-election-seeking members). Similarly, legislators who are retiring from office but not termed-out are also less likely to propose particularistic legislation, initiating about three fewer targeted bills and about six fewer local bills than their re-election-seeking counterparts. ¹⁶ Members who are termed-out yet pursuing another elective office, however, do not exhibit significantly different behavior from legislators remaining in the chamber. These results demonstrate that severing the electoral connection reduces the incentives to sponsor particularistic legislation. However, even with term limits, some members still find it beneficial to sponsor particularistic legislation, especially if they are seeking another public office. These findings are consistent with studies of the Costa Rican political system which find that legislators facing term limits continue to sponsor particularistic legislation despite the lack of electoral incentives to do so, especially if they expect particularism will help them obtain patronage appointments (Carey 1996; Taylor 1992).

Majority party members are not more likely to propose local or targeted legislation than minority party members. Moreover, we do not find that the legislators’ ideology—as measured by W-Nominate scores—influences the decision to sponsor local or targeted legislation. Additional controls in the individual-level equation for the level of district-level electoral competition, urbanization, as well as the gender, race, and seniority level of the legislator also fail to reach conventional levels of statistical significance.

Table 3. Random Coefficients Model of Legislative Particularism in the U.S. States

$\text{{\it Notes}:} ^{*}p<0.05, \; ^{**}p<0.01,\; ^{***}p<0.001.$

State-level Variables

Inclusion of the state-level contextual effects reveals important cross-state features that shape patterns of initiating particularistic legislation. The key institutional variable hypothesized to influence legislative particularism is statistically significant. The estimated effect of Inclusiveness on initiation of particularistic legislation indicates that legislators who are elected under open primaries sponsor significantly more local and targeted legislation compared to those elected under closed primaries. In examining the first differences, we find that members in open primaries are expected to sponsor about 10 more targeted bills (about nine more local bills) compared to those running in closed primaries, all else constant. Similar to our findings about district magnitude, our results are not consistent with the predictions of the classic electoral connection literature, which implies that open primaries, which confuse a focus on individualized service delivery, should be associated with less legislative particularism. The results provide strong support for our hypothesis that legislators facing a more inclusive selectorate have incentives to cultivate a personal vote through sponsorship of district-specific legislation and are consistent with the claim that the effort to reach out to more voters in more inclusive selectorates empowers local interest groups that demand particularistic policies in return for financial support.

Despite Gamm and Kousser’s (2010) finding that legislators from richer states can afford to have more of a statewide focus, we do not find evidence that the state’s per capita income in constant dollars influences particularism. Like Gamm and Kousser (2010), we examined whether legislative turnover influenced the sponsorship of local and targeted bills. We did not find a significant effect of turnover. Neither bill introduction limits or constitutional constraints wielded a significant affect on legislative attention to local or targeted legislation. The inclusion of state house turnover rates does not affect our results. Finally, we examine the effect of party competition on legislators’ initiation of targeted and local legislation. We find that the level of party competition appears to affect the content of state legislation in important ways. One-party dominance of legislatures tends to lead to a greater focus on particularistic legislation. Legislators initiation about five more targeted bills (about six more local bills) in legislatures where the majority party occupies 83% of the seats than one in which the majority party controls 55% of the seats. ¹⁷ This finding is in line with Gamm and Kousser’s (2010) findings as well as the proposition by V. O. Key that one-party politics produces factionalism and leads to a parochialism.

Cross-level Interactions

Finally, we turn to our estimates of the conditional effects of the candidate selection system on the sponsorship of local or targeted legislation by members’ electoral circumstances. While the state-level effects show that members typically sponsor more local and targeted bills in more open candidate endorsement systems, we would not expect each member to be equally affected by the candidate selection system since in the U.S. states, some members may be subject to term limits and otherwise not facing the incentives to cultivate a personal vote. Our results indicate a negative and statistically significant relationship between termed-out and inclusiveness, suggesting an attenuation of the effect of the selection system on legislators’ sponsorship of local or targeted bills for members no longer facing voters. For members who are termed-out and not seeking another public office this is expected to attenuate the effect of the candidate selection system by about five targeted bills and three local bills. Likewise, we see a conditional effect for retiring members. For retiring members, we find an attenuation of the effect of inclusiveness on particularism of about one bill. Finally, the conditional effect of members termed out yet seeking another office and the candidate selection system on particularism was not statistically significant. These results lend further support to our theory. It is important to note that the additional cross-level interactions placed in the model as control variables (inclusiveness x majority party status and majority seat share x majority party status) are not statistically significant and exclusion of these cross-level interactions does not affect any of our other results discussed above.

Robustness Checks

We subject our result to a number of robustness checks in the Online Appendix. Rather than including District Magnitude as an individual-level variable, we include it as a state-level indicator variable Multimember Districts (equal to “1” if a state employs multimember districts). Instead of using the Inclusiveness index, which assumes that a change from a “closed primary” to a “semiclosed” primary has the same effect as a change from an “open primary” to a blanket primary, we relax the linearity assumption and instead use separate indicator variables for the different types of primaries, using closed primaries as the baseline category. We estimated our model excluding the case of Louisiana and Georgia to investigate whether this outlier case may be driving the relationships witnessed in our analysis. Our results are robust to those changes. The competitiveness of primary elections as captured by primary turnout and margin of victory at the district level and number of candidates running in the district vary within states, which has implications for legislative particularism. To address this, we include these additional controls. We also control for the size of the state house and Speaker’s powers. Our results are not affected in any meaningful ways. Additionally, we find that the number of candidates competing in the primary has a significant positive effect on the number of targeted and local bills members sponsored. The margin of victory has a significant negative effect. We also estimate models where the dependent variable is (1) the percentage of local bills sponsored in the state house and (2) the percentage of targeted bills sponsored in the state house. The results of the aggregate analysis are consistent with our other results.

We also estimated various other count models including the hierarchical Poisson regression model, the hierarchical zero-inflated Poisson regression model, and the hierarchical zero-inflated negative binomial regression model, and a hierarchical logistic regression model, in which the dependent variable was simply whether or not a legislator sponsored a local or targeted bill. The results of these models were also consistent with our main results. This gives us further confidence in the results of our models.

CONCLUSION

This work contributes to the efforts to create a general theory of legislative elections and policy-making that spans a variety of institutional settings around the world. We “blend” the electoral connection/personal vote rational choice perspective of contemporary American politics with a “voter-group alignment perspective” drawing on the comparative literature. We argue that in the context of U.S. state legislatures, where many issues involve salient concerns of distinct geographic subconstituencies within the district, legislative politics deviates from that theorized for congressional politics. We see legislators not simply as individualistic actors, as in the classic Mayhew/Fiorina framework, but as co-partisans capable of collaboration in particular electoral settings, in particular, MMDs. Most importantly, we see society as a dense set of groups with distinct interests sufficiently diverse so that distinct geographically defined groups and other interests can be divided among and targeted by different representatives within a district. Our theory predicts, contrary to the electoral connection/personal vote literature, that greater district magnitude and more open primary procedures should be associated with more legislative particularism. Using original data, we test our theory and subject our results to numerous robustness checks. Our findings provide strong support for the proposed voter-group alignment model and challenge implications of the classic personal vote/electoral connection literature in the context of state legislatures.

We acknowledge that we do not directly test the microfoundations of our theory. While the empirical patterns we identify are consistent with the implications of our theory, we do not provide any direct evidence that group voting is occurring in MMD states or states with open primary procedures. Nor do we provide any evidence that co-partisans in MMDs coordinate their electoral strategies. What we focus on is providing conclusive evidence on the effects of electoral rules and, namely, district magnitude and primary type on the tendency to sponsor particularistic legislation. We believe that our results are strong enough to raise new questions and to suggest an agenda for future research, such as for instance, case studies of the politics of MMDs to determine whether evidence exists that demonstrates that such electoral strategies exist in MMDs. While this question has largely been neglected, findings in recent research are consistent with the theory we propose. Jung, Kenny, and Lott (1994) and Schiller (2000, 115) find that the same state senators, as representatives of dual-member districts, typically divide up the district into subconstituencies along usually geographic, or economic, or other dimensions, depending on the structure of organized interests within the constituency. Similar to their counterparts in other multiseat settings around the world, senators cultivate unique policy specializations, legislative agendas, and bailiwicks, and spend less time targeting voters in neighborhoods dominated by their district mates (Schiller 2000, 114–5). The geographic basis of the personal vote is also reinforced by ballot design in some states with MMDs. For instance, ballots used in elections for the Vermont House of Representatives list candidates’ hometowns below candidates’ names. Most relevant to our analysis, based on personal interviews with legislators, Taylor (2013) argues that district mates in MMDs coordinate their electoral campaigns and collaborate to meet the demands of their constituents, working together to further the interests of different groups within the district. This allows them to reduce the cost of campaigns and avoid costly electoral competition. In line with Schiller’s findings, representatives in MMDs share in interviews that they divide up issue areas, in certain cases using geography as the basis for the divisions (Taylor 2013, 50–1) and specialize to take advantage of their unique backgrounds, and devote more time to their chosen specialty (p. 35). These findings provide some support for the causal mechanisms we propose. We believe that these and our findings warrant further research of those issues and an elaboration of the theoretical understanding of electoral rules, legislative politics, and policy-making.

This research has some broader implications. Our results beg the question: If in the context of state legislative politics U.S. voters are similar to voters in Europe or other electoral settings around the world in that they are capable of group oriented and geographically specific voting, might these same voters engage in more group-oriented voting than the electoral connection literature would suggest? Recent evidence suggests that legislators in Congress are influenced by subconstituency opinion (Bishin 2009). The empirical patterns we identify show that this issue merits further investigation. Our results also indicate that the sponsorship of particularistic policies is to a large extent a function of electoral institutions. They suggest that national policy reforms aiming at curbing targeted spending and legislative particularism more broadly should consider the effects of electoral procedures. We should point out that we focus on policy initiation. Whether policy proposals actually are implemented is a different question, which necessitates additional research.

SUPPLEMENTARY MATERIALS

To view supplementary material for this article, please visit http://dx.doi.org/10.1017/S0003055416000228