Timing Their Positions: Cosponsorship in the State Legislature

Abstract

Legislators must decide when, if ever, to cosponsor legislation. Scholars have shown legislators strategically time their positions on salient issues of national importance, but we know little about the timing of position-taking for routine bills or what this activity looks like in state legislatures. We argue that legislators’ cosponsorship decision-making depends on the type of legislation and the partisan dynamics among the current cosponsors. Members treat everyday legislation as generalized position-taking motivated by reelection, yet for key legislation, legislators are policy-oriented. With a new dataset of over 73,000 bills introduced in both chambers of the Texas state legislature in the 75th to 86th regular sessions (1997–2020), we use pooled Cox proportional hazard models to evaluate the dynamics of when legislators legislate, comparing all bills introduced with a subset of key bills. The results show that legislators time their cosponsorship activity in response to electoral vulnerability, partisanship, and the dynamics of the chamber in which they serve.

Introduction

On Monday, February 25, 2019, Senator Joan Huffman, a Republican lawyer from Houston, Texas, representing the 17th district, introduced SB21, a bill raising the minimum age to purchase tobacco products from 18 to 21 years old. The very next day, four Republicans and four Democrats joined her as cosponsors on the bill. On Wednesday of that week, Eddie Lucio, Jr., a Democrat representing the Rio Grande Valley, became the ninth cosponsor of the bill. On Tuesday of the following week, José Rodríguez (D-29) announced his cosponsorship of the bill, and Borris Miles (D-13) joined him the next day. Six days later, Beverly Powell (D-10) reported that she, too, would cosponsor. Twenty-two days after Powell’s announcement, and 37 days after Huffman introduced the bill, two more Democrats cosponsored SB21, and the following week, José Menéndez (D-26) joined the list of supporters as the final cosponsor. After a successful vote in both chambers, Republican Governor Greg Abbott signed the bill on June 7, 2019 (see Figure 3 for a graph of the timing of this bill). What explains the decision of four fellow Republican legislators to sign onto the bill so quickly and seven Democratic members to move individually, and sometimes more slowly, to declare their support?

Of their many legislative activities, roll-call voting is the most common form of position-taking, in part because legislators must take a position when a bill is up for a vote.¹ While roll-call voting forces legislators to go on the record, cosponsoring legislation is powerful because it is both voluntary and electorally meaningful. It is also a collaborative process through which legislators shape and promote policy (Bratton and Rouse 2011; Holman and Mahoney 2018).

Most previous research on cosponsorship focuses on why legislators cosponsor. But for legislators, when to cosponsor legislation is also a critical decision. In this article, we ask why legislators take positions when they do. When legislators cosponsor is important for several reasons, “some of which have to do with policy implications, others of which have to do with political implications” and all of which motivate legislators (Shipan and Shannon 2003, 656, emphasis in original). What we know about the timing of position-taking comes from Congress, and those studies are exceedingly narrow. Most often, this research examines an individual high-profile piece of legislation in one chamber of Congress at a single moment in time (Boehmke 2006; Box-Steffensmeier, Arnold, and Christopher 1997; Caldeira and Zorn 2004; Glazer et al. 1995; Huang and Theriault 2012). While important contributions, these analyses may not be generalizable beyond the unique characteristics of the individual bills.² The time is past due for a more generalizable analysis of when legislators legislate. State legislatures offer a new venue to test our strategic legislating hypotheses.

In this article, we ask “why do legislators cosponsor some bills quickly but wait days or months to cosponsor others?” We take a big picture view of strategic policymaking by examining the universe of bills, and compare that with key legislation introduced in both chambers of one legislature over 20 years. We argue that a crucial aspect of the cosponsorship decision is the choice of when to cosponsor. Our theory asserts that legislators are intentional about timing their legislative activities.

Using an original dataset on the timing of cosponsorship for over 73,000 bills introduced in the Texas state legislature from the 75th to 86th regular sessions (1997–2020), we evaluate the dynamics of when legislators legislate. Texas is a useful state to examine for several reasons. As the third most populous state, it has a semiprofessional legislature, common among several states, and is diverse in terms of geography, race, and socioeconomic status. Over the past decade, it has become more partisan, and is currently one of 38 states dominated by a single party. Texas also offers incredible transparency in its legislative actions, making comprehensive data about when legislators cosponsor bills available to citizens and researchers that other states do not. The insights from our study shed light on how electoral and partisan context, as well as the policy context, influence cosponsorship decision-making.

The results of our analyses demonstrate that legislators strategically time their cosponsorship activity in response to electoral, institutional, and partisan cross-pressures. The earliest cosponsors shape if, and when, later legislators cosponsor, but the type of legislation matters. For everyday bills, legislators cosponsor earlier when there is already meaningful bipartisan cosponsorship. However, for major legislation, legislators cosponsor earlier when more of their fellow partisans have already done so. These results highlight the importance of shifting analyses away from focusing solely on the legislative dynamics of major legislation.

Why legislators cosponsor

Scholars have long studied why legislators sponsor legislation. Many find that electoral incentives influence legislators’ behavior, including observable measures of productivity, such as roll-call votes, committee work, and cosponsorship (Fouirnaies and Hall 2022). Most legislators’ primary and proximate goal is reelection because it “must be achieved over and over if other ends are to be entertained” (Mayhew 1974, 16). Even policymakers driven to shape public policy must first gain or retain elected office (Fenno 1977).

While much of the cosponsorship and electoral connection literature begins with Congress, there is some reason to believe that elections may affect legislators’ behavior differently in state legislatures. For example, Rogers (2017) looks at whether voters hold state legislators accountable for their roll-call votes and finds no evidence of electoral accountability. Fouirnaies and Hall (2022) reach the opposite conclusion, finding that elections do influence how state legislators allocate their time. From the perspective of effectiveness, Miquel and Snyder (2006) establish that effective legislators have higher reelection rates and are less likely to be challenged. And, Garcia and Sadhwani (2022) find that some state legislators are responsive to undocumented immigrants, even without electoral incentives.

Legislators’ activities may not be directed only toward constituents. Interest groups also carefully monitor cosponsorship to inform decisions on allocating endorsements, donations, and volunteers (Arnold 1990; Bianco 1994; Mayhew 1974). Interest groups and the political parties pay attention to the amount of committee work legislators do, which may affect fundraising efforts and voter support (Fouirnaies and Hall 2018). During state legislative races, opponents highlight high absenteeism rates from roll-call votes, especially on close or high-profile votes; Brown and Goodliffe (2017) conclude that legislators are more electoral than policy-minded.

An election’s competitiveness may also affect legislators’ activity. A series of articles examined the rise in incumbents’ electoral advantage in US congressional races (see, e.g., Ferejohn 1977; Mayhew 1974), yet others dispute that accounts arguing these so-called “vanishing margins” have not made incumbents feel safer (Jacobson 1987). Regardless of how objectively competitive an election is, candidates treat each election as if it is competitive (Lazarus and Steigerwalt 2018; Mann 1978). Comparing state legislators who seek reelection and those who do not, Fouirnaies and Hall (2022) conclude that the mere threat of reelection affects legislators’ actions, even in low-salience environments such as state legislative races.

Given the importance of reelection, legislators take positions on important issues in their district to build trust among their constituents (Fenno 1978; Mayhew 1974). As Kessler and Krehbiel (1996, 555) explain, “a legislator seeking reelection could not hope for a more efficient way to strengthen his or her electoral connection than to cosponsor constituency-prized legislative initiatives.” Members only cosponsor between 3% and 5% of bills in Congress, suggesting that policymakers are selective and that it is not a “costless” activity (Bernhard and Sulkin 2013; Krehbiel 1995).

Vulnerable legislators may also cosponsor more often to gain favor with their reelection constituency (Campbell 1982; Koger 2003; Rocca and Sanchez 2008), although cosponsoring legislation can have drawbacks if the legislators’ constituents do not approve (Mayhew 1974). Challengers or the media may amplify when a member cosponsors the “wrong” legislation (Desposato, Kearney, and Crisp 2011). Overall, legislators would “prefer to have more rather than less legislation bearing their name” (Franzitch 1979, 420).

Cosponsorship is especially important for institutionally disadvantaged legislators — such as those in the minority party or representing minority communities—who have fewer opportunities to claim credit for favorable roll-call votes (Rocca and Sanchez 2008). As one member of Congress explained, “When you’re in the minority, the bills that you introduce and cosponsor define your philosophy … If you want to show your constituents who you are and what you really want to do to move the country forward, you’ve got to cosponsor bills” (Koger 2003, 232).

Beyond signaling, cosponsoring legislation can help legislators achieve their policy aims even if the bill fails (Koger 2003). Sponsoring and cosponsoring may increase the likelihood that their bill is incorporated into subsequent legislation (Bernhard and Sulkin 2013; Kingdon 1984; Koger 2003). Legislators may also cosponsor bills with the primary goal of stopping another bill (Koger 2003).

In short, the cosponsorship scholarship focuses on why legislators cosponsor legislation, that is, to achieve reelection and policy goals. What is left unanswered is when those legislators cosponsor. This leaves an important gap in the literature because cosponsorship does not happen randomly.

When legislators cosponsor

The research on when legislators take policy positions is far more limited. Most of what we know about the timing of position-taking comes from only a handful of studies, most of which evaluate timing on very notable pieces of legislation in Congress: position-taking announcements in the US House on support for the North American Free Trade Agreement (NAFTA) (Boehmke 2006; Box-Steffensmeier, Arnold, and Christopher 1997), the impeachment of President Clinton in the House (Caldeira and Zorn 2004), House veto overrides (Glazer et al. 1995), and immigration policy in Congress (Huang and Theriault 2012). Extending this logic to state legislatures is not straightforward, and we must piece together expectations from multiple sources.

First, many of the key findings are based on analyses of salient, national policy issues. For instance, Box-Steffensmeier, Arnold, and Christopher’s (1997) study of NAFTA and Huang and Theriault’s (2012) examination of the Comprehensive Immigration Reform Act find that members of Congress announce early positions based on constituency factors. Kessler and Krehbiel (1996) take a different approach and study cosponsorship timing on 51 bills with at least 50 cosponsors in the 103rd session of the US House. They find that ideological extremism shapes early position-taking more than electoral vulnerability or tenure in office. There is good reason to believe that these studies, which focus on major legislation, may not be generalizable beyond the unique characteristics of the individual bills and a snapshot in time.

Second, legislators balance several factors while considering when to announce their position (Box-Steffensmeier, Arnold, and Christopher 1997; Kingdon 1989). As Caldeira and Zorn (2004, 526) explain, “To the extent that position-taking represents a very public act on the part of a legislator, the decision of when to engage in such behavior assumes strategic importance.” If position-taking was not strategic, legislators would join on as cosponsors en masse. Instead, we assert that legislators carefully weigh the costs and benefits of when to cosponsor. And, each choice—to delay or to act—is part of a member’s legislative calculus. Announcing a position too early may carry political risk, but delaying may also have potential costs. Based on these electoral, institutional, and partisan cross-pressures, we expect legislators to be strategic about when they choose to cosponsor legislation.

Legislators may cosponsor early to shape the future debate, to establish issue ownership or expertise, or to signal strong support for, or opposition to, an issue. For example, taking a position early establishes a legislator as a policy entrepreneur and signals to colleagues, especially to copartisans, the position to take (Box-Steffensmeier, Arnold, and Christopher 1997; Lohmann 1993). Legislators must also evaluate potential reactions to each bill—miscalculating may mobilize a challenger in the next legislative race (Schiller 1995)—so a member may take extra time before announcing a position (Lazarus and Steigerwalt 2018). They may also wait until outside political actors, such as interest groups or the executive branch, become involved in the issue, or to entice a concession from party leadership (Boehmke 2006). Late position takers may also respond to colleagues’ earlier positions and cues (Krehbiel 1991; Lohmann 1993).

Given the thousands of bills introduced in a legislative chamber in a session, policymakers use heuristics, or information shortcuts (Kingdon 1989; Matthews and Stimson 1975), to inform their decisions about what and when to cosponsor legislation. Many legislators take cues from experts (Fong 2020), and may have to wait for or seek their advice. Legislators may delay taking a position on an issue, even until the presiding officer announces a roll-call vote, to avoid constraint and tension from various constituents, to receive more information, or to increase the probability that their vote is vital to the outcome of a bill (Box-Steffensmeier, Arnold, and Christopher 1997; Glazer et al. 1995; Zelizer 2019).

The number and distribution of cosponsors on a bill can say a lot about the bill’s content, popularity, and the ease with which it will sail through the chamber. Thus, legislators spend considerable energy recruiting cosponsors and building coalitions, sometimes compromising on key provisions or trading votes (Bernhard and Sulkin 2013; Koger 2003). Having many cosponsors signals to colleagues the broad appeal of the proposed legislation (Bernhard and Sulkin 2013).³ As the number of cosponsors increases, policymakers may be reasonably assured that the bill has wide popularity and decide to cosponsor earlier (but see Zelizer 2019, finding that cue-taking occurs late in the policymaking process). Legislators perceive bills with many cosponsors as more likely to pass and may want their name on “winning legislation” for the potential electoral gain (Franzitch 1979). As legislators establish the appeal of the bill, legislators may find “safety in numbers” and feel comfortable making an earlier cosponsorship decision. Thus, we expect legislators use the number of cosponsors as an indicator that the bill has broad appeal and is not controversial. As a result, we expect:

Bandwagon Hypothesis: As the total number of cosponsors increases, legislators will cosponsor earlier relative to other bills.

Legislators look to their copartisans for guidance on their cosponsorship decision-making calculus (Kessler and Krehbiel 1996; Kingdon 1989). Members with a similar ideology and partisanship are likely to hold similar policy preferences. Legislators collaborate with the legislators with whom they are in policy agreement (Swift and VanderMolen 2016). As Trubowitz and Mellow (2005, 435) note, legislators “avoid taking policy positions that might antagonize party activists, campaign contributors, and core supporters.” Legislators’ partisanship and the need to gain and maintain support from partisan voters structures their policymaking activity (Aldrich 1995; Cox and McCubbins 1993), who see bipartisanship as a policy loss for their own party (Harbridge, Malhotra, and Harrison 2014) and reward roll-call vote extremism (Birkhead 2015). The nature of polarization matters; as polarization between Democratic and Republican women in the chamber increases, members are more likely to introduce restrictive abortion bills (Matthews, Kreitzer, and Schilling 2020). Bipartisan cosponsorship might marginalize a legislator if constituents perceived them to be a party traitor (Holman and Mahoney 2018). Bills with many cosponsors from one party indicate the popularity and significance of the bill within the party, also signaling it will be favorable to party donors. A cosponsorship coalition of members primarily from one party may encourage members from that party to cosponsor earlier. We expect:

Copartisan Hypothesis: As the number of cosponsors from within their political party increases, legislators will cosponsor earlier than bills with fewer copartisan cosponsors.

While some view partisan loyalty in cosponsorship as positive, others praise bipartisanship, and there is some evidence of bipartisanship in state legislatures (Kirkland 2014; Makse 2020). Having bipartisan cosponsorship can suggest that the framing and content of the bill is amenable to both parties (Crisp, Kanthak, and Leijonhufvud 2004; Koger 2003). Party leaders use the number and diversity of cosponsors in their calculations of the time and effort needed to pass legislation; when legislators who typically disagree come together to support a bill, it suggests that the bill is uncontroversial and thus relatively low cost (Koger 2003). Thus, a slate of bipartisan cosponsors may reveal broad support and increase the chances of passage (Holman and Mahoney 2018; Kanthak and Krause 2012; Kirkland 2011). When the cosponsors of a bill come from both parties, legislators may find that there is less electoral risk in supporting the legislation, and thus be eager to sign on as a cosponsor earlier.

However, not all bipartisanship represents meaningful collaboration across parties. A bill with 60 cosponsors but with 59 from the same party may technically be bipartisan, but not in the genuine spirit of bipartisanship. Being the sole cosponsor from one party may put one at great electoral risk of being labeled a party defector (Holman and Mahoney 2018). However, when bipartisanship moves beyond token status, legislators may be more likely to see little risk in joining and thus cosponsor earlier. We expect:

Bipartisan Hypothesis: When there is substantial bipartisanship (each party has at least 20% of the total cosponsors), legislators will cosponsor earlier than bills with fewer bipartisan cosponsors.

Legislators who are electorally vulnerable cosponsor more than their colleagues, as it provides a venue for taking more policy positions (Burkett and Skvoretz 2005; Garand and Burke 2006; Rocca and Sanchez 2008). But these electorally vulnerable legislators may not cosponsor earlier. Legislators take early political or policy positions when it is less electorally risky for them (Caldeira and Zorn 2004; Glazer et al. 1995). If position-taking is an activity undertaken by legislators in pursuit of reelection, then legislators who ran uncontested in their previous general election likely feel less pressure to sponsor legislation that will gain them favor with their constituents. State legislative elections have long exhibited low rates of contested elections, particularly in states with a dominant party (Burden and Snyder 2021; Myers 2020). When legislators do not face electoral challengers, they may be slower to cosponsor. In contrast, legislators who face competitive elections may cosponsor earlier in their haste to take many positions. Thus, we expect:

Election Hypothesis: Legislators who had a challenger in the general election will cosponsor earlier than legislators who did not face a challenger.

Scholars have long focused on major legislation, anomalous by definition,⁴ relative to average or symbolic bills (but see notable exceptions, including Dodd and Schraufnagel 2009; Durr, Gilmour, and Wolbrecht 1997; Howell et al. 2000). With the watchful eyes of the media on debates over key pieces of legislation, it is reasonable to assume that high stakes and public attention shape cosponsorship. In contrast, the vast majority of legislation exists in a low-information environment, gets negligible or no media coverage, and is not a critical issue in electoral campaigns. In low-information environments, partisan cues prevail (Peterson 2017). As such, we expect that legislators will time cosponsorship differently based on whether it is a major piece of legislation.

Everyday and Key Legislation Hypothesis: Partisan cosponsors will have a stronger influence on the timing of cosponsorship decisions of “everyday” bills, while substantial bipartisanship will more strongly influence the timing of “key” legislation.

Finally, we expect there to be meaningful differences between the upper and lower chambers of the legislature. Institutional factors shape cosponsorship, such as chamber size, term limits, and electoral considerations associated with shorter terms in office (Kirkland and Williams 2014; Rippere 2016; Sarbaugh-Thompson et al. 2006). Shorter terms in office, term limits, and chamber size in the lower chamber influence opportunities to foster collaborative relationships across the aisle, making partisan cosponsorship less likely. Institutional mechanisms in the upper chamber, namely the filibuster (through which individual senators wield significant influence), may induce more bipartisanship than the lower chamber (which is more majoritarian in nature).

Cosponsorship in the Texas state legislature

We examine position-taking in the Texas state legislature for many reasons. First, Texas has a semi-professional or hybrid legislature, where legislators meet for regular sessions beginning in January of odd-numbered years and meet for 140 days. Because the legislature is part-time and legislators earn a salary of approximately $45,000 for a two-year term,⁵ most legislators work elsewhere with careers as varied as an orthopedic surgeon, accountant, banker, and owner of an insurance agency (Sandoval, Montgomery, and Fernandez 2021).

The Texas legislature is a relatively large legislature with 150 House and 31 Senate members.⁶ It follows a similar process to ones used in many legislatures across the country. For example, like Texas, most state legislatures have the filibuster (in the Senate), they list one author as the “primary author,” and there can be an unlimited number of cosponsors. Like many other states, there are no term limits. The institutional differences between the chambers are like most state legislatures: longer terms for Senate members than House members (four-year versus two-year terms), staggered elections in the Senate, and a significantly smaller number of members in the Senate.

As the third most populous state in the nation, Texas is diverse in terms of geography, race, and the socioeconomic status of its districts. Like many state legislatures, Texas has a one-party regime in which the Republican Party dominates state politics. Texas has been a “state government trifecta” or single-party government with Republican control of both legislative chambers and the governorship since 2003. State trifectas are common. In 2021, there were 38 states with trifectas: 15 states with Democratic trifectas and 23 with Republican trifectas.⁷

We also examine cosponsorship in Texas for pragmatic reasons. The state provides significant transparency regarding each step of its legislative process, including the dates that legislators sign on to the bill. It is extremely rare to find the dates of cosponsorship, much less for all legislation over time. The comprehensive and longitudinal nature of the Texas legislative data opens the door to unique analyses that simply are not possible in other legislatures.

To test our expectations about the strategic legislating hypotheses, we require data on all bills introduced and legislators’ cosponsorship activity.⁸ The Texas legislature identifies one legislator as the primary author, and up to four additional joint authors. Members can cosponsor bills in the other chamber, but we do not use that information for this analysis (for cross chamber cosponsorship in the 81st session, see Kirkland and Williams 2014). What most scholars would think of as “cosponsoring,” Texas calls “coauthoring.⁹ To be consistent with state legislative literature, we use the language of cosponsoring.

To construct the dataset, we scraped all regular bills introduced in both chambers of the Texas state legislature from the Texas Legislature Online website.¹⁰ The data starts at the beginning of the 75th regular session in 1997, the first year that complete cosponsorship data is available, through the 86th regular session that ended in 2020. This process produced 73,458 bills.

We retrieved several pieces of useful information for each bill. First, we collected the roster of what Texas refers to as primary bill authors, joint authors, and cosponsors and the dates on which legislators announced their position. We also collected key information, such as the bill number, a bill description, committee assignments for both chambers, and the last legislative action for each bill.

Figure 1 reports the total number of bills introduced and the total number of cosponsors in each regular session by chamber. Total cosponsorship in both chambers experienced an increase in the later years. Figure 2a presents the average number of bills each legislator cosponsors in a session. Mean cosponsorship follows a similar pattern to total cosponsorship, with both the Senate and House’s mean cosponsorship increasing. Figure 2b shows the average number of cosponsors per bill, which remains relatively low given the total number of bills introduced each session. These averages do not tell the full story. Some legislators do not cosponsor any legislation, and others cosponsor a significant number of bills. For example, Senator Judith Zaffirini, a Democrat who has served in the Texas Senate since 1987, cosponsored at least 100 bills in each legislative session, despite serving in the minority almost the entire time of her service.

Figure 1. Total cosponsorship in each regular session.

Figure 2a. Mean number of bills cosponsored.

Figure 2b. Mean number of cosponsors.

In order to test the differences between all legislation and major legislation, we collected information on Texas’s major legislation. We do this by looking at those bills considered a “key vote” by the experts at Vote Smart.¹¹ Vote Smart recruits experts from each state (a community of political scientists and journalists) to evaluate which proposed bills were important that year. The experts evaluate the votes based on a variety of criteria, including whether it was a salient issue, a close vote, simple to understand, or definitive in establishing a legislator’s position on a particular issue.

Vote Smart data is not available for the entire period; however, it covers a robust time span, from 2009 to 2020. There are 133 key vote bills, with 63 bills in the House and 70 bills in the Senate.¹² Notably, not every key vote bill becomes law. The legislation covers a wide range of topics, including health care, environment, immigration, voting rights, and the budget.

Methodological approach and dependent variable

Since we are interested in estimating the timing at which legislators announce their positions, we use duration analysis. Specifically, we use a pooled proportional hazards Cox model (Cox 1972), which is a flexible duration model that leaves the distribution of the baseline hazard unspecified (Box-Steffensmeier and Jones 2004).

Our dependent variable is the number of days it takes for a legislator to cosponsor each bill per session. We operationalize it as a count of the number of days it takes for a legislator to cosponsor a bill after the first person sponsors the bill.¹³ In event history analysis language, the legislator “survives” until the date she cosponsors a bill, the bill is passed, or the session ends, whichever comes first. She “fails” when she signs on as a cosponsor of that bill. We can use the anecdote at the beginning of this article as an example to illustrate our dependent variable: the 2019 Senate bill SB21 about raising the minimum age to purchase tobacco products. As shown in Figure 3, a Republican senator introduced the bill. This is day 0. For SB21, day 0 is February 25, 2019. On February 26, eight legislators cosponsored the bill. For each of these eight legislators, their clock stopped on day 1. When Borris Miles cosponsored the bill on March 6, his clock ended after 9 days. The final senator to cosponsor SB21, José Menéndez, cosponsored on April 9, the 43rd day. For the remaining senators who have not cosponsored the bill, their clock ends either when the bill passes the chamber or when the session ends, whichever occurs first. The timing for SB21 and all other bills that are in that session are independent of one another in this type of analysis.

Figure 3. 2019 Senate bill SB21 cosponsorship timing.

The Cox duration model has several advantages. First, the dependent variable calculates not only if a legislator cosponsors, but when they do. Second, the Cox model explicitly analyzes legislators’ dynamic deliberative process. For example, the models do not have a constant term because it is absorbed into the baseline hazard rate.

Finally, duration models are particularly well suited to handle censored data. In our analysis, observations are right-censored, which occurs when the session ends if a legislator never cosponsors. To account for this, duration models create a variable to specify whether a legislator never cosponsored (data is right-censored) or cosponsored legislation (data is not censored). Including the censored data is important to ensure that we account for all legislators’ actions and their timing. Excluding the right-censored data produces selection bias and loss of information (Box-Steffensmeier 1996). Including censored data does not cause parameter estimate bias because the censored observations contribute to the likelihood function, which indicates the probability of surviving, that is, not cosponsoring until time t (Yamaguchi 1991).

The “pooled” component of the model means that we place all legislators and bills per session together in one dataset. In our first set of models, we pool across all bills and all sessions for each chamber. This means that we have an observation for every legislator, for every bill introduced in their respective chamber, from the 75th session in 1997 to the end of the 86th session in 2020. For the second set of models, we still pool across sessions (81st to 86th) but only focus on the key vote bills identified by Vote Smart.

Because we pool the data, it is important to be clear about what we are modeling. The unit of analysis is the legislator, per bill and session. That means that for each bill per session, every legislator is “at risk” of cosponsoring that specific bill. If she does not cosponsor the bill, the clock ends on the last day of the session and the case becomes right-censored. In Texas, legislators introduce a considerable number of bills that never attract cosponsors, meaning that all legislators are at risk for those bills for the entire session. For every bill introduced, there is an observation for each individual legislator, since they are all at “risk” of signing on as a cosponsor for each bill. In the 85th session in the House, for example, 152 members introduced 4,333 bills. For this session, there are 668,901 total observations.¹⁴

Independent variables: Cosponsor characteristics

Using the Legislative Reference Library of Texas website,¹⁵ we retrieved the name, partisanship, district, chamber, and incumbency status of every legislator from 1997 to 2020, from which we generate many of our independent variables. First, we focus on the variables of interest for this analysis—the size and partisanship of the cosponsors of the bill. We first construct a variable that keeps a running tally of the number of cosponsors a bill has at the time each legislator signs onto the bill. The second variable of interest we include is the number of co-partisans that have signed onto a particular bill. If the legislator is Republican, this variable will be the number of Republicans who had already joined the bill at the time of the signing. If the legislator is a Democrat, it will be the number of Democrats.

The final variable centered on the partisan makeup of the cosponsors considers the amount of bipartisan support there is for a bill. In our theory, we argue that as cosponsorship moves from trivial bipartisanship to substantial, other legislators will be more likely to cosponsor. We create an indicator variable where the variable equals 1 if at least 20% of the cosponsors are from both parties, and 0 if not.¹⁶ We chose 20% because it represents a meaningful proportion of bipartisanship according to the data. We include a variable to indicate whether the legislator had a challenger in their previous general election (Klarner 2018; Klarner et al. 2013, supplemented by our own data collection). In the House, 40% of the legislators, on average, did not have a challenger in their previous election. Senators experienced low levels of competition (sometimes reaching 50% of the senators) in the early years of the data, but in more recent years, the average has dipped to about 20%.

To account for a legislator’s ideology, we use the American Legislatures project NPAT scores (Shor and McCarty 2011). NPAT scores are individual-level ideal point estimates that place all legislators on a common ideological space, so the scores are comparable across time and between and within state legislative chambers. The larger the positive value in these scores, the more conservative a legislator is. A large negative number reveals that a legislator is very liberal. To control for how extreme a legislator is, we take the absolute value of the NPAT ideology measure.

Independent variables: Constituent characteristics

Legislators also time their positions based on constituent factors (Box-Steffensmeier, Arnold, and Christopher 1997). There is a relationship between cosponsorship and constituent issue preferences (Bromley-Trujillo, Holman, and Sandoval 2019; Waggoner 2018) A legislator whose ideology closely matches the preferences of her constituents will have little concern for taking an early stance on an issue compared with a member whose district is more ideologically heterogeneous or less ideologically proximate. To account for citizen ideology, we use Tausanovitch and Warshaw’s (2013) district-level ideology measure.¹⁷ We use citizen and legislator ideology to categorize the similarity between the representative and their district.

To do that, we split each of the ideology measures into three sections—the most liberal third, the moderate third, and the most conservative third. We then calculate the absolute difference between the legislator’s position and their district’s position. If a legislator’s ideology was in the most conservative third of the chamber and their district was one of the more liberal districts, the difference would be 2. These scores range from 0 to 2; in both chambers and across all sessions, we find that the mean is 0.45. Although not perfect, there is no current state politics measure that places citizens and legislators in the same space. This gives us a general sense of how similar legislators are to their constituents.¹⁸

Independent variables: Additional control variables

Using the data scraped from the Texas Legislature website, we construct several additional variables to control for bill characteristics. Since we know the primary sponsor of each bill, we include the ideology of the bill using a proxy: the primary sponsor’s NPAT score, offering a measure of the bill’s conservativism. To control for the substantive focus of the legislation, we break down the committee assignments into 13 categories, which often include multiple committees. The categories are Agriculture, Business, Defense, Education, Energy, Environment, Health, Justice, Parks and Recreation, Taxation and Revenue, State and Local Government, Welfare, and Transportation. Transportation is the omitted base category from all models. In both chambers, there are several bills that are never assigned to a committee, although the number is much larger in the Senate than in the House.

Results

We present two series of models.¹⁹ In the first, we analyze every bill introduced in the House and the Senate from 1997 to 2020. Next, we limit our analysis to key pieces of legislation. For all results, we report hazard ratios; a coefficient greater than 1 means that a legislator cosponsors earlier and a coefficient that is less than 1 indicates that a legislator delays cosponsorship. We run all models with robust standard errors, stratified by legislative session. Finally, Cox models assume that the covariates’ effects are constant over time, known as the “proportionality assumption.” We test whether the models violate this proportionality assumption using Schoenfeld residuals. Overall, most of the covariates meet the assumption.²⁰

In Table 1, we present the results for the House and the Senate for all bills. The results support our strategic legislating hypotheses, although there are some chamber differences. As discussed above, differences in chamber size, length of terms, and institutional rules (like the presence of the filibuster) may shape how and when legislators collaborate. As the number of cosponsors increase in the House, legislators cosponsor earlier. For example, as the number of cosponsors grows from 1 to 3, other legislators in this House are over five times more likely to cosponsor earlier. But the results are in the opposite direction in the Senate. As the number of overall cosponsors grows in the Senate, legislators delay cosponsorship. For a similar change in the number of cosponsors (from 1 to 3), there is a comparable decrease with the decision to cosponsor—they are almost five times less likely to sign on to the bill.

Table 1. Cosponsorship timing for all legislation, 1997–2020

Note. Coefficients are hazard ratios, a coefficient greater than 1 means that a legislator cosponsors early, and anything less than 1 means that a legislator cosponsors later. Outcome is the number of days until a member cosponsors each bill.

* p < 0.05.

^** p < 0.01.

^*** p < 0.001.

The next two variables look at the party of the cosponsor. First, in both chambers, as the number of copartisans increases, legislators cosponsor much earlier. As the total number of copartisan cosponsors changes from one to three copartisans, legislators are 8 times more likely in the House and 10 times more likely in the Senate to cosponsor earlier.

Second, bipartisanship is a significant predictor of cosponsoring earlier. The very large, positive, and statistically significant hazard ratios indicate that as bipartisanship exceeds 20%, legislators cosponsor earlier. This means that the probability of cosponsoring legislation with significant bipartisan support is over 15 times that of bills without meaningful bipartisan support. Bipartisanship is by far the most effective factor in changing the likelihood that a legislator will sign onto a bill.

Finally, electoral vulnerability affects the likelihood of cosponsoring a bill. Having a challenger in the previous elections makes a legislator cosponsor earlier in the House, but has the opposite effect in the Senate. This may be because of the senators’ longer terms. In the Senate, having a challenger in the previous election decreases a legislator’s hazard ratio by over 100% versus not having a challenger. In the House, having a challenger increases the hazard ratio by almost the same amount. The relationship between electoral pressure and the use of cosponsorship is stronger in the House versus the Senate.

Our first analysis examined every bill introduced. While that analysis offers insight into general trends for the universe of legislation, legislators may time their cosponsorship differently on especially salient issues. To analyze whether there is a distinction between types of legislation, we next turn to the Vote Smart key vote legislation.

Table 2 illustrates that legislators time their cosponsorship a bit differently for major legislation. As the total number of cosponsors increases, legislators delay their cosponsorship of key vote bills. The substantive effects are similar in the House and the Senate; as the overall number of cosponsors increases from 1 to 3, legislators will be five times more likely to delay cosponsoring the bill.

Table 2. Cosponsorship timing for key vote legislation, 2009–20

Note. Coefficients are hazard ratios, a coefficient greater than 1 means that a legislator cosponsors early, and anything less than 1 means that a legislator cosponsors later. Outcome is the number of days until a member cosponsors each bill.

* p < 0.05.

^** p < 0.01.

^*** p < 0.001.

The copartisan effect remains the same. As the number of copartisan cosponsors increases, legislators will cosponsor earlier. This effect is stronger in the Senate than it is in the House for both key vote and everyday legislation. In the House, increasing the number of copartisan cosponsors on the bill by one increases the likelihood of announcing one’s cosponsor by over 1,000%, whereas in the Senate, a similar increase increases the hazard ratio by over 7,000%. Interestingly, the powerful effect of bipartisanship on the universe of legislation disappears for key vote bills, but only in the Senate. Meaningful bipartisan cosponsorship is statistically significant in the House model. Finally, electoral vulnerability does not influence when legislators cosponsor key votes bills.

Discussion

One of the most surprising results from the models—the universe of bills and key vote bills—is that more cosponsors on a bill does not mean legislators cosponsor earlier. As the total number of cosponsors increases, legislators delay cosponsorship. We expected a general bandwagon effect in which legislators jump on early because everyone else is cosponsoring. Instead, legislators’ timing is more nuanced than that. For major legislation, a partisan bandwagon effect emerges. As the number of cosponsors from within their own political party increases, legislators cosponsor earlier. The effect is strongest in the Senate. However, for typical or everyday legislation, legislators cosponsor earlier when existing cosponsorship is bipartisan.

Legislators’ ideology sometimes affects cosponsorship decision-making. Across all models and in both chambers, ideologically extreme legislators are more likely to cosponsor early. This is as expected because these members often represent more ideologically extreme districts and can cosponsor early without fear of being out of step with their constituents.

In addition, electoral vulnerability only influences cosponsorship decision-making for everyday bills. The typical legislation may provide legislators more opportunities to stake their claims in policy areas, whereas major legislation may make it harder for legislators to do the same. We also find that there is a difference between how electoral vulnerability changes the decision-making process across chambers. In the lower chamber, legislators are much more likely to sign on early when they have faced a recent electoral challenger, but that legislators in the upper chamber who faced a recent challenger delay cosponsorship activity.

There are mixed results when we account for the ideology of the bill author. In the House, members cosponsor earlier when the bill’s author is more conservative, especially when the bill is a major piece of legislation. But the effect is not the same in the Senate; members only cosponsor earlier in the Senate on major legislation. Perhaps because it is Texas, using the ideology of the primary author as a proxy, the key vote bills are more conservative on average than the everyday bills. In the House, the average ideology score is 0.1488, a fairly moderate score, whereas the key vote bills have an average of 0.574, better reflecting the Republican majority’s policy agenda. The differences in the Senate are much more drastic and may explain why there are no statistically significant findings in the everyday legislation. For the Senate, the average ideology score is only 0.096, a nearly perfect moderate position, whereas their key vote bills are strongly conservative at 1.072. As the bill author becomes more conservative, Senate legislators delay cosponsoring on key votes. Finally, the ideology of the district does not affect cosponsorship timing; it is not statistically significant in most models.

Partisanship affects cosponsorship decision-making in other ways. Copartisanship is a reliable predictor of cosponsoring early for key votes and for the universe of bills. Bipartisanship, however, does not reach statistical significance for key votes. Democratic legislators have mixed results based on the type of legislation. Among all bills, Democrats will cosponsor earlier. However, for key votes, it depends on the chamber. In the House, Democrats delay their cosponsorship, but there is not a statistically significant effect in the Senate. This difference is likely because of institutional differences. There are nearly five times as many members in the House as the Senate, so more bills originate in the House and members have more opportunities to cosponsor. Together, the partisan variables reveal the majority party’s control of the agenda. For the most important legislation, Republicans do not need to rely on Democrats to advance their agenda, and Democrats will probably not receive much benefit in cosponsoring Republican bills, and so delay their cosponsorship activity.

Conclusion

Our work offers several key contributions. First, moving the analysis from the US Congress to the state level, while also including all bills introduced, allows us to identify different timing strategies for key and everyday legislation. Our findings reveal that legislators strategically time their cosponsorship in response to electoral vulnerability, partisanship, and the dynamics of the chamber in which they serve. In response to the previous work on US congressional position-taking, we add legislators are strategic in state legislatures, even with non-contentious, regular policymaking.

Together, the results reveal that there are two types of cosponsorship: generalized and policy-focused. Legislators treat most bills introduced as generalized position-taking, in which reelection is their primary objective, and thus cosponsorship serves as a relatively costless, but nontrivial, signal to their constituents. But not all legislation is the same. For key votes, legislators employ a different cosponsorship decision-making calculus. Given the high-profile and influential nature of these bills, members look beyond reelection toward actual policymaking. Here, Republicans do not need to gather bipartisan cosponsors, or even to accumulate many cosponsors, because they are in the majority in both chambers. Instead, they focus their efforts on fellow partisans to ensure that they have the votes to pass the legislation they want. Electoral vulnerability has no influence over the decision to sign onto a key bill, but has important effects on this calculus for all bills.

A final important contribution involves the data. First, our analysis examines all bills, providing a more generalizable picture of position-taking. Second, the universe of legislation examined includes bills in which there was no cosponsorship. This is equally important because the absence of cosponsorship also affects cosponsorship. In looking at routine policymaking, we show that many legislators do not take a position at all. In fact, some legislators cosponsor only a few bills and some never do. As a result, cosponsorship is a relatively rare phenomenon in the data.

While the Texas legislature is not unique in having an abundance of bills with few cosponsors, analyzing data with many zeroes can by challenging. Despite this, our results are robust to different model specifications, including using Democrats as the base category instead of Republicans, using split sample based on legislator partisanship, alternative measures for variables such as ideology, the inclusion or exclusion of committees, and controlling for chamber leadership in the second set of pooled models. We also ran models specifying certain covariates as time-varying and tested other methods for handling tied failures.

As a first look at the timing of regular, non-salient position-taking, these analyses resolve some questions but raise additional avenues for future research. In this paper, we took a direct approach to studying policymaking by looking at all proposed legislation, regardless of many potentially relevant factors, such as the bill’s content, technicality, or salience, which future research should explore. Yet another avenue to pursue would be examining legislatures with more party competition as well as how factions within the dominant party inform strategic cosponsorship decisions. Future work may also explore to what degree legislators cosponsor certain bills based on the intersection of their race, gender, and partisanship.