Progress or Principle? Partisan Competition, Bill Sponsorship, and Position-Taking in Congress

Abstract

This paper demonstrates how partisan competition over majority control of Congress influences the viability of legislators' lawmaking activities. More specifically, I develop a dynamic pivotal politics model of policy change, delineating the conditions under which partisan agenda-setters will respond to competition over majority control by slowing policy change, discouraging members from expending effort to draft viable, compromise legislation. I then test the predictions of this model using an original set of spatial point estimates for status quo and bill proposal locations in Congress, based on co-sponsorship and interest-group position-taking data. Using these data, I find strong support for my model's predictions. In particular, I find that members of Congress are far more likely to offer messaging bills when the theory suggests party leaders will block otherwise viable legislation, for partisan competitive reasons. The findings speak to a growing literature tying the insecurity of legislative majorities to a wide variety of legislative outcomes.

Full text

Progress or Principle?
Partisan Competition, Bill Sponsorship, and
Position-Taking in Congress
——–
Jesse M. Crosson
Princeton University & Trinity University
November 6, 2019
Abstract: is paper demonstrates how partisan competition over majority control of
Congress influences the viability of legislators’ lawmaking activities. More specifically,
I develop a dynamic pivotal politics model of policy change, delineating the conditions
under which partisan agenda-setters will respond to competition over majority control
by slowing policy change, discouraging members from expending effort to draft viable,
compromise legislation. I then test the predictions of this model using an original set of
spatial point estimates for status quo and bill proposal locations, based on co-sponsorship
and interest-group position-taking data. Using these data, I find strong support for my
model’s predictions. In particular, I find that members of Congress are far more likely
to offer messaging bills when the theory suggests party leaders will block otherwise viable
legislation, for partisan competitive reasons. e findings speak to a growing literature
tying the insecurity of legislative majorities to a wide variety of legislative outcomes.
1

Aside from voting, bill sponsorship is among the most fundamental behaviors in which a legislator
may engage. Indeed, before a legislature ever considers a bill for passage, lawmakers and their staff
must first draft it, often making difficult decisions about which provisions to include or exclude,
based on policy goals and the prevailing political climate. Members exercise this care for good reason:
presumably they want to draft a bill that can pass. Indeed, sponsoring successful legislation redounds
to the member’s benefit in numerous ways, even beyond policy gains: legislative successes generate
opportunities for credit-claiming, and they raise a member’s profile among her peers.
In spite of the centrality of this goal, members do sometimes sponsor legislation that they understand
will not pass. Progressive members, for example, introduced bills to implement a single-payer health
care system in 2009. Similarly, some Democratic members have recently introduced bills that would
abolish Immigration and Customs Enforecment (ICE)—despite the fact that Republican leaders were
highly unlikely to move on such bills. Further still, in nearly each year since its passage, Republicans
have introduced legislation meant to fully repeal the Affordable Care Act, while other conservative
Republicans have even sponsored bills to abolish the IRS in recent Congresses.1
Given members’ positions as lawmakers, the introduction of such nonviable legislation is puzzling, at
least from a policymaking perspective: why do members expend effort in drafting bills they understand
will not become law? In the abstract, previous literature has offered a plausible baseline explain for
the value of such bills to members: sometimes termed “messaging” bills, nonviable legislation offers
members the opportunity to position-take before key electoral constituencies. Still, in spite of the
importance of this baseline explanation, previous research has not provided a means for understanding
the trade-off between position-taking and policymaking that members face when drafting legislation.
In this paper, I argue that such decisions are a function of members’ reactions to the institutional
constraints and electoral climate surrounding them. More specificatlly, I argue that electoral competition
over major institutional pivots sometimes encourages members to carefully draft viable legislation with
the highest probability of passage, while it at other times discourages members from doing so. Given
that individual members understand when party leaders face incentives against setting the legislative
agenda, members respond rationally by adjusting their bill sponsorship strategies, understanding the
wastefulness of drafting costly viable bill-drafting under such conditions—instead offering nonviable
1Such bills have become so common, in fact, that some research has even referred to them as “dead on arrival” bills
(Gelman, 2017).
2

“messaging” legislation.
To demonstrate how electoral context and congressional institutions influence members’ willingness
to draft viable legislation—legislation that, if afforded agenda space would pass through Congress—I
develop and test a bill-level theory that ties traditional spatial models of policy change with the differential
policymaking incentives encouraged by the recent rise in competition over majority control of Congress.
e theory demonstrates how expectations over future control of Congress influences party leaders’
willingness to set the legislative agenda for status quo policies lying within certain subsets of the policy
space, which in turn influences the kinds of bills members introduce for these status quos. Using a
new dataset of point estimates for both status quo and bill proposal locations derived from a joint
scaling of cosponsorship, roll call, and interest group position-taking data (and generated on the same
scale as preference estimates for members of Congress), I test and find support for the theory’s bill-level
predicitions—namely that members draft viable legislation when electoral incentives and the location of
a bill’s associated status quo encourage party leaders to set the agenda, and resort non-viable messaging
bills (bills that do not improve upon the status quo for key veto players in Congress) otherwise.
Even beyond understanding why and when legislators engage in earnest lawmaking, these findings
point to the policymaking ramifications of heightened competition over control of a legislature, which
enables agenda-setters to consider post-electoral dynamics in the first place. Much like its level of
preference polarization, Congress’s level of competitiveness has fluctuated dramatically over its history.
is study ties members’ sponsorship activities to this competition, demonstrating that changes in
Congress’s competitiveness may have far-reaching consequences for member behavior. Along the way,
the study also makes use of valuable data that allow for the a priori measurement of status quo locations
in Congress—even when bills do not ultimately receive a roll call vote. While previous research by
ieme (2018) has previously developed such measures using interest group, cosponsorship, and roll
call data at the state level, the tests in this paper are the first to make use of such data at the federal level.
I proceed as follows. First, I review relevant literature on sponsorship activity, demonstrating the
need for a better understanding of both the electoral foundations of bill sponsorship activity in general
and the determinants of whether members sponsor viable legislation in specific. Second, I develop
a theory of the underlying agenda-setting process that bills face, given various electoral expectations
faced by partisan leaders. e theory suggests that when partisan agenda setters believe a status quo
policy is better moved after the upcoming election, legislators are more likely to introduce non-viable
3

messaging bills. ird, I detail the data and measurement strategy used to test the theory’s specific
hypotheses regarding electoral expectations, the spatial locations of bills and status quo policies, and
the introduction of viable legislation. Finally, I provide empirical evidence in support of my theory,
showing that the agenda-setting incentives generated by partisan electoral context appear to influence
how and when members offer viable or messaging legislation. I conclude by discussing the study’s
implications for the study of legislative behavior.
Bill Sponsorship, Position-Taking, and Electoral Competition
Traditionally, research on bill introductions has conceptualized sponsorship as a tool for achieving
policy goals. Wawro (2000), for example, features bill sponsorship prominently in his examination
of legislative entrepreneurship in the U.S. House. Similarly, Volden and Wiseman (2014) incorporate
a member’s bill sponsorships directly into their measure of legislative effectiveness. Such a focus makes
sense, given the lawmaking responsibilities of members of Congress; consequently, most examinations
of bill sponsorship feature explanatory variables situated within the policymaking process rather than in
the electoral process or in communications. Such factors include majority status, committee membership
and status, proximity to key institutional pivots, and investment in legislative staff (Schiller, 1995;
Garand and Burke, 2006; Cox and Terry, 2008).
Still, members of Congress do occasionally employ primarily legislative behaviors for non-legislative
purposes. Hall (1996), for instance, argues that member participation in committee is not uniformly
motivated by a desire to influence policy outcomes. Instead, some “showhorse” members use the
committee as a means for magnifying their communications efforts and better position themselves
for re-election. More specific to bill sponsorship, Sulkin (2005) finds that politicians’ promises on
the campaign trail translate to actual sponsorship activity: when politicians make promises to address
particular issues while campaigning, they frequently sponsor related legislation once they take office.
ese findings provide some context for the otherwise puzzling observation that members occasionally
draft legislation that is not politically viable and would not likely pass through Congress even if voted
upon. at is, apart from policymaking, such findings underscore that bill sponsorship offers position-taking
value to the member. In fact, Rocca and Gordon (2010) show that members frequently use bill
sponsorship as a means for public position-taking, especially before interest groups. Yet while position-taking
value provides a rationale for why members expend effort drafting nonviable legislation, few studies
4

have offered a theory for why and when the position-taking value of bill sponsorship predominates
over its policymaking value.
Understanding the conditions under which sponsorship serves policy-change versus position-taking
goals is important for a wide variety of reasons. First and foremost, most scholars consider policymaking
to be the primary representational function that members of Congress serve in American democracy.
Insofar as members expend valuable time and resources drafting legislation that stands little chance of
passing, one must wonder how it affects their ability to discharge other key duties of the office. Second,
while bill sponsorship is frequently incorporated into measures of legislative effort and effectiveness,
the presence of non-viable legislation should count for less in such measures than bills carefully crafted
to maximize chances for passage. Moreover, given that bill sponsorship factors prominently in some
measures of legislator homestyle, understanding how bill sponsorship is used for legislative versus
non-legislative purposes once again proves integral to accurately capturing home style among modern
legislators. Finally, in an era of insecure majorities (Lee, 2016), both individual members and parties
have increasingly emphasized messaging over policymaking, in an effort to maximize reelection chances
and seat share in Congress. Consequently, understanding how and when members deploy messaging
legislation is central to capturing how increases in competition over majority control have altered legislative
activity within Congress.
Below, I develop a bill-level theory of legislative viability, which links this contemporary rise ini
insecure congressional majorities to members’ bill-specific decisions to pursue different types of bill
sponsorships. More specifically, I offer a spatial theory of policy change that delineates how agenda-setters
in Congress may strategically speed up or slow down changes to specific status quo policies, based
on their party’s anticipated electoral gains or losses in the coming election. ese agenda-setting
expectations, then, determine whether members are willing to expend the effort necessary to draft viable
legislation. Using a new dataset of bill proposal locations and their associated status quo locations, I
show that when members expect agenda-setters to slow down the policymaking process for a specific
status quo location, they are less likely to meet that status quo with a viable proposal—that is, a proposal
that, on the basis of its spatial location improves upon the status quo for all relevant veto points in
Congress. Conversely, when members expect agenda-setters to accelerate the policymaking process for
a specific status quo location, members meet such status quos with viable proposals that improve upon
the status quo for all congressional veto players. Put differently, they offer legislation that, if brought
5

up for a vote, should be expected to pass through Congress.
A Theory of Policy Change and Sponsorship Type
Typically, theories and empirical examinations of bill sponsorship focus solely on members’ decisions
regarding whether or not to sponsor legislation at all. at is, based on the member’s own goals, or
her issue commitments during the campaign (Sulkin, 2005), a member simply selects which status
quo policies to target and sponsors legislation accordingly. Here, I argue that members make an
additional decision when sponsoring legislation, deciding whether or not to invest in write a viable
legislative proposal that could conceivably pass into law. at is, while the decision of whether to
sponsor legislation is largely a function of campaign promises and other electoral considerations made
by the member at the beginning of her term in office, her decision of how seriously to pursue these
lawmaking goals hinges on her expectations of whether agenda-setters will move on proposed changes.
In determing whether to pair specific status quo policies with viable or nonviable proposals, I
argue that members weigh how electoral and institutional dynamics influence the probability that their
bill will receive agenda space. us, to understand when members should offer viable legislation, I
develop a dynamic spatial theory that delineates when agenda-setters are likely to speed up or slow
down the policymaking process. In short, I show that when partisan electoral expectations encourage
agenda-setters to slow down the policymaking process, members face little incentive to pay the steep
costs associated with drafting viable legislation. Conversely, when electoral conditions encourage
agenda-setters to speed up the policymaking process, members face a greater incentive to sincerely pursue
their lawmaking goals. To more precisely articulate this dynamic, I develop a unidimensional spatial
theory of agenda-setting and policy change with an endogenous (between rounds) status quo and two
actors, an agenda-setter (AS) and a receiver (R).
Electoral Expectations and Partisan Agenda-Setting Decisions
Before members decide on their sponsorship strategies for a given status quo policy SQi, they observe
the following agenda-setting game. In the first or “present” round, the agenda-setter (AS) must decide
for SQiwhether or not to propose an alternative a(“propose” versus “hold back”).2If she does propose
2Note that status quo policies are indexed by i, in order to indicate that the agenda-setter encounters many status quo
policies within a given legislative period.
6

an alternative, the game shifts to the receiver (R), who must then select whether to “accept” or “reject”
a. If the receiver chooses to accept the alternative, the game ends, the equilibrium policy SQ∗=a,
and payoffs are realized via a quadratic loss function comparing the new policy to each of the players’
ideal points. Should the receiver choose to reject the alternative, the status quo SQ persists. us, the
result of Round 1 can be either a new policy or the status quo, much as in any traditional spatial model.
Unlike traditional spatial models, however, if SQ is reached because of “holding back” behavior
by AS, the game does not end.3Instead, an election occurs, shifting the location of agenda-setter to
AS′and the receiver to R′.4In the second round, the game proceeds as in Round 1: AS first decides
whether or not to propose an alternative to the status quo, and R decides whether to accept or reject
that proposal. If the proposal is accepted, the game ends with a new policy of SQ∗
i=a. If the proposal
is rejected, the game ends with the same status quo policy, SQ∗
i=SQi.
A key feature of this game’s structure is the fact that the game only reaches the second round if the
status quo persists—a feature designed to replicate realistic trade-offs that agenda-setters face within
issue areas. Substantively, this structure creates a key decision for AS: she must choose between what
she believes she can gain by proposing a new policy in this round, versus what she believes would occur
(for each SQi) following the next election. e rationale behind this feature is drawn from substantive
observations of the American legislative system: when policy change occurs for a status quo policy in
the present legislative session, it is highly unlikely to occur again in the next session. Policy change
for a specific status quo policy area can either occur now or later, but not both. Policy advocates and,
increasingly, political researchers (e.g., Buisseret and Bernhardt (2017)) denote this feature of legislative
politics frequently, and the interested reader may find further justification for this structural decision
in Supplemental Information E.
Key Model Features
Members of Congress observe these agenda-setting dynamics before deciding whether to pair a particular
SQiwith a viable or non-viable proposal. However, before discussing how such agenda-setting dynamics
3Technically, the game could reach the second round if Rrejects AS’s offer a. However, because Ris not dynamically
sophisticated and AS knows R’s preferences, AS chooses not to make offers in the first round that she knows will not be
accepted, assuming an infinitely small proposal cost.
4Notationally, then, if a shift in AS or R does occur in Round 2, I will refer to said second-round actors as AS′and R′.
If, however, no change occurs, I will simply refer to AS and R similarly in both rounds.
7

influence members’ sponsorship decisions, I first underscore some key features of the agenda-setting
model as presented above.
First, while previous models of policy change, such as Krehbiel (1998), include a larger number of
players with specific identities, I keep the number of players (and the specificity of those players) low,
in order to increase the flexibility of the theory. at is to say, because the specific identities of the
agenda-setter and pivotal actor are fluid, subject to intense scholarly debate, or some combination of
the two, this design simplifies the bargaining environment into two key players—an agenda-setter and
a single veto agent—who do not have to take on any specific identity. Doing so allows one to make a
specific theoretical point and extend to a variety of political contexts.
In order to eventually interpret the dynamics of the theory in the American context, however, I do
include a few key assumptions with regard to player locations and identities. More specifically, I assume
that agenda control rests in political parties, much in the same way as Cox and McCubbins (2005).
With respect to the receiver, I rely upon the fact that, in one dimension, a single actor is ultimately
pivotal for any particular legislative decision. In this model, then, Rlies at the pivotal actor located
farthest from AS in the opposite ideological direction (rightward if AS is leftist, leftward otherwise).
is setup generates an asymmetry between AS and R. at is, while AS is a collective actor (a party,
or set of party leaders) that persists across elections, Ris most often an individual legislator, concerned
about reelection. is asymmetry has important consequences for how the two players approach the
game. Given that party leaders are likely to remain in place (either as minority or majority leaders)
following the next election, they are enabled to think dynamically about policy change and weigh the
advantages and disadvantages associated with proposing or holding back various pieces of legislation.
Consequently, AS is a dynamically sophisticated player who considers Period 2 consequences to its
actions in Period 1.
is conceptualization of AS stands in sharp contrast to R. As an individual actor, Rfaces the
real possibility that she may not remain in office following the upcoming election. If the current actor
Rfails to remain in office, then future policy gains are of little use to her. Put differently, insofar as
reelection concerns remain as the individual legislator’s primary concern (Mayhew, 1974), dynamic
considerations regarding policy gains likely fall to the wayside. erefore, given that Rfaces such
pressures in the present round (pressures that AS does not face, at least at the same level), Ris modeled
as a “static” player in the game. In other words, R votes in accordance with her present-round incentives,
8

accepting policy proposals that move the status quo in her direction and rejecting those that do not.
It may initially seem tempting to think of Rin terms of a dynamically sophisticated minority party:
if the majority party is dynamically sophisticated, why is the minority party not thinking dynamically
and whipping Raccordingly? To be clear, it is in fact likely that minority parties are dynamically
sophisticated in some sense: minority parties want to maximize their chances of taking back the
majority. However, even if they were able to whip moderate members located near the Rpivot, doing
so is not likely to improve the minority’s chances at taking the majority. To see why, consider what
might happen if Rdid vote dynamically—in other words, to occasionally vote for policies that move
the status quo away from his/her ideal point or against policies that move the status quo closer. Such
votes are not likely to improve the chances that the minority party regains power: voting against her
own preferences (and potentially the preferences of the district) is unlikely to improve R’s reelection
chances. Understanding this, marginal members occasionally respond quite colorfully to the prospect
of their being whipped in this fashion. For example, moderate Sen. Joe Manchin (D - WV) recently
took umbrage at the idea of Senate Minority Leader Chuck Schumer influencing his vote: “I’ll be 71
years old in August, you’re going to whip me? Kiss my you know what.”5Assuming that the minority
party does not view losing incumbent seats as a viable strategy for regaining power, then, it may choose
against cross-pressuring Rin her vote choice. Rendering Ra static player thus makes sense, even if one
thinks of her as under the power of the minority party.6
e analysis will rely upon a few other assumptions worth mentioning. First, the model setup
implies that the agenda-setter can never be “crossed” by the receiver, if/when the receiver moves in the
direction of the agenda-setter. Operationally, this simply means that Rlies to the right of a left-leaning
AS and to the left of a right-leaning AS. Additionally, I assume that AS and Rcannot share an ideal
point.ese are weak assumptions that improve model interpretation.
How Do Electoral Prospects Influence Policy Change Under Possible
Power Transitions?
Given the structure and features of the agenda-setting game, what sorts of agenda-setting dynamics
do members of Congress observe as electoral dynamics change? As is common for spatial models,
5https://www.politico.com/story/2018/07/13/schumer-supreme-court-fight-centrist-democrats-716654
6It is worth noting that, practically speaking, this feature mirrors an assumption made by Buisseret and Bernhardt (2017)
in their recent paper.
9

the equilibria of this game vary considerably with regard to status quo location and the locations of
AS, R, AS′, and R′. us, to illustrate the sorts of agenda-setting behavior members of Congress
should expect for a given SQi, I detail how three power transition scenarios influence AS’s willingness
to set the legislative agenda.7Much like Krehbiel (1998) and others, my discussion of the theoretical
dynamics underlying these regimes will remain abstract, for illustrative purposes; regardless, the scenarios
underscore conditions under which members of Congress can expect specific SQito be met with
accelerated or decelerated agenda-setting (terms I define below), which influences members’ willingness
to draft costly viable legislation.
For each scenario, I delineate the conditions under which AS ought to set the agenda, comparing
results to those from a traditional, static model. en, I develop three main predictions regarding
agenda-setting activity, highlighting when party leaders will speed up or slow down policy change.
Finally, I trace out relevant implications for bill sponsorship activity from these predictions, ultimately
arguing that future gains for the majority party encourage sincere policymaking, while future losses
may discourage it.
Scenario 1: AS Maintains Agenda Control and Makes Gains with Reciever
In the first power transition scenario, AS is expected to maintain agenda control following the upcoming
election. However, Ris expected to move closer to AS . Here, without loss of generality, suppose that
AS lies left of center and Rto the right. As noted above, the upcoming election is expected to be a
positive one for AS’s party: in addition to retaining control of AS,Ris expected to move closer to
AS. at is, by making gains in R’s legislative chamber (and, say, capturing the filibuster pivot, for
instance), AS and her party expect to experience a closer R′after the upcoming election.8Should the
expected shift in R’s location occur, players in the game have a reasonably reliable idea about where the
new receiver, R′, would be located.
How does this possibility of change of control influence AS’s actions in the present round? Consider
how AS ought to act if the probability of R moving closer is equal to 1 (Pr(R′−AS < R −AS) = 1).
In this scenario, AS must backward induct from the second round, to determine where policy would
7I eventually argue that all post-WWII elections each fall into one of these regimes.
8As Figure 1 depicts, AS’s retention of agenda control is captured by the persistence of AS in the second round. at
is, the location of AS in the second round is equivalent to that in the first round. In reality, this is unlikely to be the case. If,
for example, Democrats add seats to their majority, the location of AS is likely to shift slightly leftward. However, because
such intraparty shifts are typically small, AS is held in place here, for ease of exposition.
10

Figure 1: Common Power Distribution and Electoral Change Scenarios; Scenario 1
R
R’
AS
AS
SCENARIO 1: AS Does Not Shift and R Moves Closer to AS
1
A
2
AS
SCENARIO 2: AS Does Not Shift and R Moves Farther from AS
1
2
SCENARIO 3: AS Shifts from Democratic to Republican Control, Thereby Moving R
1
2
AS
R’
R
AS
AS’
R
R’
move should she opt against offering a policy alternative in the present round. Consider first a status
quo policy lying far to the left of AS. For such status quo policies, AS can offer an alternative policy
located at her ideal point, because such an alternative is a net improvement for the Republican receiver.
Because AS can do no better in the second round by holding back, she instead should always propose
her ideal point in the first round for any such status quo policy.
is dynamic changes for SQilying at AS and rightward. Indeed, if the status quo lies close to but
to the right of AS, AS may desire to move the status quo but cannot do so: Rwill reject any movement
away from her (R’s) right-leaning ideal point. Moreover, for all policies located between AS and R′,
SQ will persist through both rounds, as neither the first-round nor second-round agenda-setter will be
able to make improvements upon the status quo. In the present round, AS will be unable also to move
status quo policies between R and R′in her direction, so her best response is simply to allow the status
quo to persist.
However, for policies lying to the right of R, AS faces an interesting incentive. Policies lying to
the right of R are moveable in the first round: R will accept any proposal at least as good as the
11

status quo. However, for these SQ to the right of R, SQ is even less desirable for R′than it is for
R. Consequently, AS can extract more policy concessions in the second round than the first. Taken
together, AS faces an incentive to hold back from offering a policy alternative when the status quo is to
the right of R, even though she can improve upon the status quo in the first round by making an offer.
is dynamic is not limitless, however. Indeed, eventually a status quo policy is so far to the right that
AS ≥SQ −2|SQ −R|—i.e., that AS’s ideal point lies within the leftward reflection of SQ over
R. Under such conditions, AS can propose and obtain her ideal point in the first round, rather than
having to wait until the second round. is means that for any weak preference of present gains over
future ones, AS should propose her ideal point in the first round, which R will accept.
e results are summarized in the upper portion of Figure 2. Here, the horizontal axis represents the
location of the SQ, while the vertical axis represents the location of the equilibrium policy outcome,
SQ∗. e dark line tracks the equilibrium outcome SQ∗for each SQ along the horizontal axis. Finally,
the gray portion of the graph covers the region of SQ values for which no policy change occurs in the
first round of play. As Figure 2 depicts, policy stasis occurs not only for SQ between AS and R, but
also for policies lying to the right of R.9is region is quite large, indicating that policy change should
slow significantly when AS expects to face a more favorable Rin the future. Compared to the static
mode (bottom panel of 2), the dynamic model predicts considerably less agenda-setting and eventual
policy change.
ese results hold for any scenario under which AS remains the same and makes some kind of gains
with the receiver in the upcoming election. Under these scenarios, less policy change will occur than
what models based on static preferences alone would predict: for SQ policies lying to the right of
R,AS is at best indifferent between Round 1 and Round 2 policy outcomes, opting against Round 1
policy change for all policies to the right of Rand to the left of R+|AS −R|. In empirical tests in later
sections, I call this phenomenon policy deceleration. is phenomenon is summarized in Proposition 1:
Proposition 1 (Policy Deceleration):AS will refrain from attempts to change status quo
policies lying to the right of Rbut to the left of R+|AS −R|when AS anticipates her
party will remain in control of AS and Rwill move closer to her.
9Here again, it is worth noting that while the equilibrium outcome (SQ∗=AS) is unambiguous for policies lying
to the right of the reflection of AS over R, whether or not such change occurs in the first or second round depends upon
assumptions about temporal preferences on the part of AS.
12

Figure 2: R′More Favorable to AS
AS
R
R’
SQ*
AS R’ R
“Present” round gridlock interval
Equilibrium outcome
Scenario 1 Dynamic Game
AS
R
R’
SQ*
AS R’ R
Hypothetical Round 2 outcome
R + |AS –R|
Scenario 1 Static Game, Before and After Election
In other words, when AS realizes she can achieve a more favorable change to SQiafter the upcoming
election, she will elect to postpone movements of SQiafter the election.
Bill Sponsorship Under Scenario 1
Clearly, at least in the abstract, dynamic electoral considerations may dramatically influence how agenda-setters
think about the legislative agenda. Given that the game detailed above is one of full information,
it stands to reason that the game’s dynamics should also influence how members think about bill
sponsorship. Indeed, when a member finds herself in a situation like Scenario 1 and wishes to address
a particular status quo by introducing legislation, she does so with an understanding that AS faces
incentives to decelerate policymaking for certain regions of the policy space.
How might individual members respond to such conditions of policy deceleration? Consider the
costs associated with the two types of bill-writing discussed above: viable and non-viable sponsorship.
As noted throughout, a piece of viable legislation is a bill that should pass into law, should it be brought
up for a vote. Within the context of the spatial model, viable legislation must be spatially acceptable to
all pivots or veto players: that is, viable legislation must serve as an improvement upon the bill’s associated
status quo for all veto players within the political system. Such legislation therefore moves the status quo
13

toward the center of the political spectrum (relatively speaking). Conversely, non-viable legislation is
not an improvement for one or more veto players, meaning the sponsor has elected to move the status
quo away from the center of spectrum—typically close to their own ideal point.
In order for members to draft truly viable legislation, they therefore must compile a large amount
of political and policy-specific information. Indeed, beyond grasping the legal, economic, and social
ramifications of various policy instruments, a member must explore how pivotal legislators and interest
groups are likely to react to policy proposals. Compiling such information is costly, occupying a sizable
portion of a member’s time and legislative resources. Ultimately, the member does receive a benefit
from sponsoring this type of legislation: should the bill pass, the related policy gains would benefit
her. Moreover, she may gain the respect of her colleagues, and she may be viewed as productive by her
constituents. Still, because she will likely need to compromise from her preferred policy outcome, her
position-taking payoff with her reelection constituency is limited.
By contrast, non-viable bills do not require the member to compromise on her preferred policy
outcome. Indeed, sometimes termed “messaging” bills, the primary purpose of such bills is to offer
an opportunity for a member to signal her alignment and commitment to the ideological principles
of her reelection constituency (Gelman, 2017; Rocca and Gordon, 2010). Unlike viable legislation,
the member recognizes that messaging bills are not likely to pass into law, even if they do receive a
vote in Congress. Consequently, the member need not expend valuable legislative time and resources
compiling information about key legislative actors. Instead, she need only ensure she maximizes
position-taking benefit from sponsorship such legislation. us, while she forfeits potential policy gains
by offering messaging legislation, she scores political points with key supporters, potentially aiding in
her reelection.
I argue that the relative value of viable and non-viable legislation therefore depends upon the
probability that movements of specific SQiwill occur, all else equal. More specifically, the value
of viable and non-viable proposals fluctuates based on the likelihood that agenda-setters will actually
move on legislation that alters a given status quo. Scenario 1 generates conditions under which viable
proposals, at least for some SQi, the development and introduction of a viable proposal makes little
sense. at is, when a status quo policy lies within the deceleration region, members should offer
non-viable or messaging legislation: because AS is unlikely to move on such legislation, paying the
legislative costs associated with introducing viable legislation makes little sense. Instead, she should
14

maximize the non-policy benefits of bill sponsorship, offering messaging legislation.
Taken together, the trade-offs associated with viable and messaging legislation therefore interact with
expectations over future electoral outcomes to translate Proposition 1 into a testable hypothesis about
bill sponsorship activity:
H1:When AS is not likely to change but Ris expected to move closer to AS after the
upcoming election, legislators aiming to change policies within the deceleratin region will
be less likely to offer viable proposals.
at is, when electoral incentives and the location of pivotal actors render a specific status quo policy
subject to policy deceleration (i.e., SQ opposite AS), members should be systematically more likely to
offer non-viable or messaging legislation, all else equal.
Scenario 2: AS Remains the Same, but Receiver Moves Away from AS
Unlike Scenario 1, AS loses ground with Rin Scenario 2. Here, AS and Rboth lie on the leftward
portion of the political spectrum; however, following the election, R is expected to move away from
AS, rendering Round 2 far less advantageous to their policymaking endeavors.
How does this potential shift influence the strategic calculus made by AS in the present?10 As in
Scenario 1, status quo policies lying to the left of the agenda-setter are moveable to AS’s ideal point in
the first round: Rwill accept any movement of these status quo policies to the right. Similarly, policies
located between AS and Rare immoveable, regardless of the location of R—meaning that the status
quo remains in place within this range. But what about status quo policies lying to the right of R? In
Scenario 1, AS opted to hold back from policy change. But unlike in Scenario 1, AS should no longer
hold back on these status quos. In fact, one might argue that AS should accelerate her policymaking
efforts on a subset of these status quos. Following the election, status quo policies lying between R
and R′become immoveable. erefore, if AS wants to lock in policy gains in this area, she needs to
propose changes now. Of course, as in Scenario 1, this dynamic is not equally true of all SQ to the
right of R: for SQ to the right of the reflection of AS over R′,AS can achieve her ideal point in either
round. Because AS cannot improve upon this outcome under such conditions, she can make an offer
in Round 1 and the game ends.
10Considering again the scenario wherein the rightward shift of R is guaranteed to occur (Pr(R′−AS > R −AS ) = 1).
15

Figure 3: AS Loses Ground with R′
AS
R’
R
SQ*
AS RR’
“Present” round gridlock interval
Equilibrium outcome
Scenario 2 Dynamic Game
AS
R’
R
SQ*
AS RR’
Hypothetical Round 2 outcome
R’ + |AS – R’|
Scenario 2 Static Game, Before and After Election
“Time sensitive” status quo
policies
Figure 3 summarizes these results and compares them to the static case. For the present-round static
case, nothing has changed: status quo policies lying between AS and R will remain unchanged. Strictly
speaking, rendering the game dynamic did not increase or decrease the number of moveable status quo
policies in equilibrium. However, the game’s dynamism accelerates policymaking in a different way
(or, at very least, focuses it): that is, because AS knows that policies between Rand R′may become
immoveable in the immediate future, she may exert additional effort in moving these policies. Whether
or not this results in more policy change overall depends on how scarce agenda space is, but the theory’s
dynamism at very least suggests where AS is likely to focus her policymaking efforts. More specifically,
AS is likely to accelerate policymaking efforts for policies located to the right of R, but to the left of
R′+|R′−AS|. I refer to this phenomenon as policy acceleration, which stands in stark contrast to the
policy deceleration encouraged in Scenario 1. I define policy acceleration as follows:
Definition 1 (Policy Acceleration): the choice by AS to focus her policymaking efforts
on a particular set of status quo policies, due to her belief that future changes of these
policies will benefit her less than current ones.
A necessary condition for policy acceleration, of course, is limited agenda space. Were agenda space
16

unlimited, AS could successfully address all moveable status quo policies, regardless of whether such
policies become immoveable following the upcoming election. Consequently, the probability that AS
will move any particular SQ would not differ other moveable SQ. With limited agenda space, however,
AS must prioritize. Given that her payoffs are a function of her proximity to the eventual SQ∗,AS can
ensure the best possible two-round policymaking outcome by focusing on SQimovements that differ
substantially between Round 1 and Round 2. For Scenario 2, these policies are located to the right of
R, where Round 2 gains are either impossible or minor. Proposition 2 summarizes this phenomenon:
Proposition 2: When AS expects to retain control of the AS position but lose ground
with R, she will focus her policymaking efforts on status quos located to the right of R
but to the left of R′+|AS −R′|.
In other words, for status quo policies that AS understands will become immoveable following the
upcoming election, she will accelerate her efforts to address those policies.
Bill Sponsorship under Scenario 2
As underscored above, members’ decisions between viable and non-viable sponsorship are in part a
function of their beliefs over the probability that their proposed legislation will actually be put up for
a vote by majority party leaders. Unlike Scenario 1, conditions of policy acceleration render payment
of viable legislation’s costs more beneficial. at is, given that legislation is more likely to move for
certain SQiwithin Scenario 1, members should capitalize by offering legislation that could pass, were it
brought up for a vote—opting against messaging for the time being. More specifically, when interested
in addressing SQilocated inside the acceleration region, members of Congress are better served to offer
viable legislation than they would be under other electoral conditions. Formally:
H2:When AS is not likely to change but Ris expected to move farther from AS after
the upcoming election, legislators aiming to change SQ within the acceleration region
will be more likely to offer viable proposals.
at is, when a member has committed to addressing a status quo policy that is subject to policy
acceleration dynamics (SQ opposite AS , under the aforementioned electoral conditions), she should
be systematically more likely to pair that SQiwith a viable—and not messaging—piece of legislation.
17

Scenario 3: Control of AS Changes
In Scenarios 1 and 2, AS is expected to remain unchanged. In these scenarios, AS therefore considers
only how changes in Rinfluence her policy change options in present and future legislative periods.
However, when control of AS is competitive, policymaking dynamics grow more complicated. Under
this scenario, control of AS is competitive: it is expected to switch from Democratic to Republican
control. Due to the definition of Rdescribed earlier, the location of Rwill therefore also change. at
is, because the identity of Ris defined as the farthest veto player from AS, a change in AS necessitates
a change in R. In this example, if AS lies leftward and Rrightward, a shift in AS will move Rleftward.
How do such major changes affect the policymaking dynamic? Consider first the status quo policies
lying to the left of AS. In spite of the potential for coming changes, AS’s dominant strategy for these
status quo policies remains unchanged from previous scenarios. Indeed, AS can achieve her ideal point
in the first round, because Raccepts any rightward movement in these status quo policies. Because
AS cannot improve upon this result, she will always offer AS in the first round for status quo policies
lying to her left.
For slightly more conservative status quo policies, however, a different dynamic begins to emerge.
Consider what might happen if status quo policies lying between AS and R′are allowed to persist into
Round 2. For these status quos, AS′may exploit the favorable Rlocation and move policy rightward by
2|SQ −R′|. is result is far worse for AS than the status quo. How, then can AS respond? Recall that
any new policy change in Round 1 ends the game for that particular status quo. Given this feature of
the game, AS can protect against rightward movements of status quo policies in this range by offering
SQ∗=S Q to R. In other words, by offering a policy that is identical to SQ,AS can achieve an
outcome that is better for herself than what would occur in the second round. While this strategy may
seem at first unrealistic, a practical application of this sort of dynamic may occur when a majority party
chooses to reauthorize a program without making major changes to the program’s structure. Instead of
allowing to the next Congress to take the reauthorization, the current agenda-setter can lock in, say, 5
more years of the current program structure and policies.
is incentive for AS to make SQ-equivalent offers disappears for status quo policies lying to the
right of R′. First, whereas policies lying between AS and R′were vulnerable to rightward movements
by AS′, policies lying between R′and Rare located within the Round 2 gridlock interval. No offer
18

Figure 4: Identity of AS Changes
“Present” round gridlock interval
Equilibrium outcome
Scenario 3 Dynamic Game
Hypothetical Round 2 outcome
Scenario 3 Static Game, Before and After Election
“Time sensitive” status quo
policies
AS
R
R’
SQ*
AS R’ RR + |AS –R|
AS’
AS’
AS
R
R’
SQ*
AS R’ RR + |AS –R|
AS’
AS’
AS could make in the present round would improve upon these status quo policies, so policy change
does not occur within this interval. Status quo policies to the right of R, however, are moveable, just
as in the static case.
Figure 4 depicts these equilibrium policymaking outcomes, SQ∗. In this case, dynamic outcomes
differ very little from those in the static case. As the gray regions of the figure demonstrate, the set of
immoveable status quo policies in the dynamic case differs only slightly from that in the static case.
Indeed, aside from the SQ-equivalent offers made within the [AS, R′] interval, there is no set of SQi
for which AS’s ability to pursue would differ from the static case. Similarly, as the dark lines clearly
demonstrate, the equilibrium policy outcomes do not differ at all between the static and dynamic cases.
ese similarities notwithstanding, much as in Scenario 2, AS does face incentives in Scenario 3
to accelerate policymaking for a certain subset of status quo policies in the first round of play. is
incentive is depicted in Figure 4. Consider first SQ policies lying to the right of R. Given that status
quo policies to the right of Rwill be either immoveable (R′< SQ < AS′) or moved to AS′(SQ >
AS′), AS should accelerate her reform of status quo policies to the right of R, out of concern for poor
Round 2 outcomes if no new policy is adopted. Her reason for doing so is clearly illustrated by the
difference between the dark and light lines found in the bottom half of Figure 4. For policies to the
19

right of R, the light line (which represents the equilibrium outcome if AS allows a given SQ to persist
to the second round) lies consistently farther away from AS on the vertical axis than does the dark line,
which represents the equilibrium outcome associated with AS deciding to change a given SQ in the
first round. As the distance between these lines grows, AS faces an increasing incentive to accelerate
policymaking in the first round.
As the distance between the light and dark lines demonstrates, incentives for policy acceleration are
also strong for status quo policies lying to the left of AS. For such policies, rather than AS achieving
her ideal point by changing such SQs in Round 1, AS′can move SQ significantly rightward. Given
that R′is, by assumption, located to the right of AS, such rightward shifts will always result in a policy
change that is worse for AS than her Round 1 outcome (her own ideal point).11 Consequently, for SQ
located to the left of AS, the discrepancy between the dark and light lines is considerable: AS stands
to lose a great deal if she fails to address SQ within this region. Given this discrepancy, AS may wish
to especially accelerate policymaking efforts for SQ to the left of AS.
e policy acceleration incentives associated with Scenario 3 are summarized in Propositions 3a and
3b.
Proposition 3a: When AS expects that her party will lose agenda control, she will
accelerate policymaking efforts throughout the policymaking space.
Proposition 3b: AS will focus her policy acceleration on SQ policies for which the
difference between aand SQ′is greatest—namely, policies located to the extremes of AS
and AS′.
In other words, a shift in control of AS generates strong incentives for the agenda-setter to accelerate
policymaking for all non-gridlocked status quos. ough not examined directly in this paper, this
incentive is especially strong for extreme status quo policies that lie opposite the agenda-setter.
Bill Sponsorship under Scenario 3
Similar to Scenario 2, the policy acceleration throughout much of the policy space in Scenario 3
generates conditions favorable to the introduction of viable legislation. Indeed, because members
11Recall that, by assumption, AS is always more extreme than R. erefore, when Rshifts toward AS, it remains more
moderate in its preferences than AS. In this case, when R′shifts to the left due to the change in AS , we can say that R′
will lie to the right of AS’s current location.
20

understand that AS faces incentives to set the agenda for bills lying outside [AS, R], they should be
more likely to pay the costs associated with viable bill sponsorship, all else equal. Here, the sponsorship
dynamics are similar to Scenario 2 in theory, though policy acceleration—and therefore, viable bill
introduction—extends throughout a much larger portion of the policy space in Scenario 3. In sum,
the dynamics in Scenario 3 lead to a third and final hypothesis:
H3: When AS is likely to change hands, legislators aiming to change SQ outside the
interval [AS, R]will be more likely to offer viable proposals.
at is, members interested in addressing status quo policies lying outside the static gridlock interval
should be systematically more likely to offer viable proposals when they expect control of AS will change
after the upcoming election.
Viable Bill Introduction: Measurement and Data
As this theory demonstrates, institutions and elections may combine to influence agenda-setting behaviors
in Congress. is behavior in turn alters individual members’ calculus regarding bill sponsorship
and how carefully they craft bills for passage. Translating this abstract account of agenda-setting and
sponsorship behavior into empirical tests within the congressional context implies several theoretical
and long-standing methodological challenges. I address these challenges in several distinctive ways,
which I discuss below. First, I delineate how I identify AS and Rwithin a particular Congress,
discussing relevant assumptions made about player identities and information within the policymaking
process. Second, I identify how I measure players’ contemporaneous beliefs about electoral prospects,
using information from electoral betting markets. Finally, I outline how I measure individual bill
introduction spatial locations and type—i.e., whether a member has elected to offer viable or messaging
legislation for a particular SQi. With these measurements, I am then able to test whether electoral
prospects do indeed influence bill introduction activity as hypothesized.
Identification and Measurement of AS, R, AS′,and R′
In order to identify the locations of AS and Rwithin a given Congress, one must first identify which
actors in Congress count as pivotal. e U.S. Constitution identifies several such pivotal actors: the
median member of each chamber of Congress, the president, and (when relevant) the veto override
21

pivot. However, legislative scholars also recognize other veto points in the U.S. policymaking system.
First, most recognize the 60th vote in the Senate—the filibuster pivot—as a veto player. Additionally,
previous studies have posited that partisan control of the voting agenda introduces additional veto
points into the political system. In particular, the application of the Hastert Rule12 in the U.S. House
renders the median of the majority party pivotal (see, for example, Woon and Cook 2015 and Crosson
2019). While some accounts of the U.S. Congress argue that similar agenda control may exist within
the U.S. Senate, considerably more debate exists on this point. Consequently, for the purposes of
this empirical analysis, I assume the following actors are pivotal in U.S. federal policymaking: both
chamber medians, the U.S. House majority median, the Senate filibuster pivot, the president, and the
veto override pivot. However, given the measurement concerns underscored below, I focus solely on
legislative veto players in this analysis, meaning AS and Rare located at the House chamber median,
House majority median, Senate chamber median, or Senate filibuster pivot.
Given that only one of these veto players—the House majority median—is an explicitly partisan
actor, my empirical analyses assume AS is located at the House majority median. Spatial models always
include assumptions about agenda power, and previous models have sometimes placed agenda-setting
power in the hands of a chamber median (e.g., Krehbiel, 1998). However, given that recent literature
finds that policy change models with partisan agenda control tend to generate the most realistic empirical
predictions, my model places this power in the hands of the majority median.
With AS located at the House majority median, Ris much easier to identify on the basis of the
model’s assumptions. at is, because Ris defined as a single actor pivotal in determining whether
AS’s proposal passes, I define Ras the veto player lying farthest away from AS. Using this set of
player-identification rules, Figure 5 outlines the locations of AS and Rover the five Congresses examined
in this study. In each of the Congresses included in this study, Rhappens to be located at the filibuster
pivot opposite AS.
Beyond measurements of AS and R, of course, calculating the acceleration and deceleration regions
requires an estimate of R′’s location. For most of the above scenarios, the precise location matters
little, if at all. For example, in Scenarios 2 and 3, players need only know, should a change in Ror
AS occur, which direction each actor will shift. For Scenario 1, however, players need to form an
12Whereby the Speaker selects to keep off the voting agenda bills that do not attain majority support from the majority
caucus.
22

Figure 5: AS and R by Congress
expectation regarding the location of R′. To generate such measurements, I lean on the assumption
that members will typically understand how electoral changes will influence who is pivotal after the
upcoming election. us, I measure the expected location of R′as the actual location of R, should the
predicted electoral outcome (discussed below) transpire.
Ideal points for these pivotal actors are measured via the preference estimates found in Crosson,
Furnas and Lorenz’s (n.d.) study on interest groups and bill locations in recent Congresses. While
the generation of these scores is detailed at greater length below, Crosson et al. generate preference
estimates for members of Congress on the same scale as a large set of point estimates for bill proposal
and status quo locations. ese preference estimates correlate strongly with existing preference measures
based on roll call data alone. For example, Crosson et al.’s preference estimates (which they call
cIGscores) for sponsors of legislation from the 110th - 114th Congresses correlate with first-dimension
DW-NOMINATE scores at ρ=.945.is correlation is particularly strong considering the fact that
the estimation matrix includes many bills, actors (interest groups), and member behaviors (cosponsorship)
not found in the matrices used to calculate DW-NOMINATE. Further still, when one compares the
members’ cIGscores with their roll-call-only IGscores, cIGscores are even more highly correlated with
previous measures, this time at ρ=.977. Figure 6 plots of sponsor cIGscores against DW-NOMINATE
scores. As with DW-NOMINATE, cIGscores remain quite bimodal in distribution and correlate
23

Figure 6: cIGscores v. DW-NOMINATE, 110th - 114th Congresses
a
aNote: Because cIGscores are static and DW-NOMINATE scores are dynamic, this graph depicts cIGscores plotted against each
legislator’s mean DW-NOMINATE score over the 110th to 114th Congress time period.
strongly with NOMINATE.
Electoral Expectations and Power Transition Scenarios
An additional requirement for testing the above hypotheses relates directly to the dynamic structure of
the game itself: in order to code whether or not acceleration or deceleration dynamics apply, one needs
to measure expectations about the upcoming congressional elections. To do so, I turn primarily to an
electoral futures market, the Iowa Electronic Market (IEM), in order to best measure contemporaneous
changes in electoral expectations—and, therefore, whether members should expect proposals to be met
with policy acceleration, deceleration, or neither.
IEM solicits “investments” from private citizens on a wide variety of political outcomes, including
whether particular candidates will win the presidency and how Senate races will end within particular
states. Most useful for this analysis, IEM has solicited wagers on partisan control of Congress since
1996. While political scientists have used these data in the past, such studies typically assess the markets
as prospective predictors of electoral outcomes—and not as a measure of how electoral expectations
24

might influence the policymaking process. However, these data are especially useful for this study’s
purposes, as they capture contemporaneous beliefs regarding potential electoral changes, rather than
actual changes observed in hindsight. Most importantly, because the contracts are priced within [$0,
$1], the resulting contract prices may be interpreted as a collective belief regarding the probability that
a chamber of Congress will be under control of either party, following the upcoming election.13
ese measurements are ideal for capturing the probability that AS will change (i.e., whether or
not the majority in the House will shift). But because Ris routinely located at the filibuster pivot,
they are less well-suited for the measurement of R’s probability for change. When the majority in
the Senate is expected to change, thereby altering the location of the filibuster pivot, the application
of IEM to measure R’s change probability is straightforward. However, if the current majoirty in
the Senate is expected to gain the 60th vote in the Senate, majority change probabilities are not
appropriate to measure the probability of change in R. In these cases, I extrapolate IEM using a variety
of contemporaneous predictors, modeling and extrapolating Senate vote share using Bayesian poisson
regression analysis. Doing so allows me to generate a posterior distribution for predicted seat shares
within each time period in my data, which I use to generate likelihoods that either party will gain more
than 60 seats. Details on this procedure are included in Appendix D.
Taken together, then, AS and Rare coded as “likely to change” according to the following rule:
CAS
j=




1if p(PAS
j̸=PAS
j)>0.5
0otherwise
CR
j=













1if p(PR
j̸=PR
j+1)>0.5or
p(V SR
j<0.6< V S R
j+1 ∨(V SR
j>0.4> V S R
j+1)>0.5
0otherwise
where CAS
jand CR
jrepresent the binary variable indicating whehter or not AS an Rare expected to
change, PAS
jand PR
jrepresent the party of AS and Rduring Congress j, and V SR
jrepresents the vote
13Unfortunately, while IEM exists for the full time period covered in this study, there are gaps in the IEM data near the
beginning of each congressional session (before new markets were opened). To address this problem, I have extrapolated the
IEM measures, using information that may inform politicians’ contemporaneous beliefs regarding probabilities of partisan
control. A more detailed recounting of this strategy is included in Appendix D.
25

share of R’s party in the Senate during Congress j. ese conditions stipulate that, if the interpolated
IEM probability (averaged over every month within a Congress) suggests that the relevant actor will
change parties with probability greater than 0.5, Cis coded as 1 (and zero otherwise). For actors located
at a chamber median, this means capturing the 50th percentile vote in the chamber; for the filibuster
pivot, this means capturing the 60th vote away from the expected AS.
If either CAS
j= 1 or CR
j= 1, then Congress jis placed into the relevant agenda-setting scenario
articulated above. More specifically:
Scenarios 1, 2: CAS
j= 0 and CR
j= 1 ;
Scenario 3: CAS
j= 1
Note that if R is projected to move closer to AS, then Scenario 1 obtains. Conversely, if R is projected
to move farther from AS, Scenario 2 obtains.
Using this measurement strategy, AS is coded as competitive in the 111th Congress (after which
control of the House switches from Democratic to Republican control), placing the 111th Congress
in Scenario 3 (policy acceleration). Ris competitive (and AS is not) in the 110th, 112th, and 113th,
and 114th Congresses. Since Rwas predicted to move closer to AS in the 110th, 112th, and 113th
Congresses, these Congresses face Scenario 1 (policy deceleration), while the 114th Congresses lies
within Scenario 2 (policy acceleration).
Measurement of Status Quo Locations and Viable, Messaging Legislation
A final requirement for testing whether bills exposed to acceleration or deceleration are systematically
more or less viable, of course, is bill-specific information regarding not only a bill’s targeted SQ, but
also the spatial location of the bill proposal itself. at is, my interest is not in measuring which SQi
the member chooses to target (which Sulkin and others have demonstrated is frequently determined
via campaign promises and is therefore exogenous to my model), but rather the seriousness with which
they attempt to change their targeted SQi. To date, widespread measures of bill proposal and status
quo locations have proven highly elusive. As Clinton (2017) summarizes in his review of strategies for
measuring the content and direction of policy changes, common methodologies for generating ideal
point estimates (e.g. Poole and Rosenthal, 1997; Clinton, Jackman and Rivers, 2004) fall short of
producing reliable estimates for prosposal and status quo locations.
26

In this study, I make use of an original dataset of congressional bill proposal locations and status
quo locations provided by Crosson, Furnas and Lorenz (n.d.)—the largest data set of congressional bill
proposal and status quo point estimates generated to date. Crosson et al. generate their estimates by
jointly scaling cosponsorship, roll call, and interest group position-taking data throughout the legislative
process. As they underscore, this approach, first developed by Peress (2013) and extended with interest
group disclosures at the state level by ieme (2018), allows for the identification of proposal locations,
cutpoints, and (consequently) status quo locations for each bill possessing the required cosponsorship,
roll call, and position-taking data. As noted earlier, the approach also generates ideal points for members
of Congress and interest groups, on the same scale as the bill proposal and status quo locations. is
information, when combined with information regarding the ideological locations of pivotal actors,
provides the final keys for the measurement of the primary variables of interests: measurements of
whether or not a bill’s associated status quo lies within the deceleration or acceleration intervals and
whether a bill is viable or messaging.
Using these data, measurement of my dependent variable, the introduction of viable or messaging
legislation, proceeds as follows. A proposal counts as viable when it would pass through both chambers of
Congress (on the basis of its spatial properties), were it to be brought up for vote—and nonviable or messaging
otherwise. us, assuming perfect information about the locations of pivotal actors, proposals are viable
when the actors located at the left and right end-points of the gridlock interval would both prefer the proposal
to the associated status quo. Because Crosson et al.’s estimation procedure generates ideal points for
members on the same scale as proposals, proposals are coded as viable if the following conditions hold:
Yijm =




1,if |SQ′
ijm −Lj| ≥ |SQij −Lj|&|SQ′
ijm −Lj| ≥ |SQij −Rj|
0,otherwise
where Yijm is the binary variable representing whether or not a member m’s bill proposal is coded as
viable, SQ′
ijm corresponds the member’s bill proposal i’s spatial location in Congress j,SQij refers to
bill i’s associated status quo, Ljrefers to the location of the most liberal pivotal actor in Congress j, and
Rjrefers to the most conservative pivotal actor. Because my theory assumes a unidimensional policy
space, if a proposal improves upon SQ for the two most extreme pivotal actors in Congress, it should
pass through Congress if brought up for a vote. Viable proposals are sensitive to these preferences and
27

improve upon the status quo for each pivotal actor.
As an example, suppose that a right-leaning member wishes to propose legislation to address a SQ
that is located to the left of [AS, R]. Suppose further that AS is also right-leaning, but that the member
in question is even more conservative than AS. If the member elects to propose a new policy at his
conservative ideal point, and if this proposal is farther away from R’s ideal point than is the SQ in
question, the proposal counts as nonviable: were it brought up for a vote, Rwould veto the legislation.
If, however, the member decides to draft legislation that moves the status quo closer to R(likely, but
not necessarily, inside of [AS, R]), the proposal would be coded as viable.
My primary independent variables, acceleration/deceleration indicator variables, are coded as follows.
First, status quo policies are subject to policy deceleration within the 110th, 112th, and 113th Congress
(given the electoral scenario codings described above) when they lie beyond Rwithin the policy space
(H1). Status quo policies are subject to acceleration in the 114th Congress when they lie beyond R
(H2). Finally, status quo policies are subject to policy acceleration in the 111th Congress anywhere
outside of [AS, R](H3). Members of Congress are expected to offer viable proposals during conditions
of policy acceleration and messaging proposals during conditions of policy deceleration. In sum:
H1:Status quo policies in the 110th, 112th, and 113th Congress lying beyond Rin the
opposite direction of AS will be met with messaging legislative proposals. Other SQ in
these Congresses will be met with viable proposals.
H2:Status quo policies in the 114th Congress lying beyond Rin the opposite direction
of AS will be met with viable legislative proposals.
H3: All status quo policies in the 111th Congress lying outside [AS, R]will be met with
viable proposals.
Additional Independent Variables
Beyond the hypothesis-specific variables highlighted above, a variety of other considerations may also
help to explain the incidence of viable and messaging proposals, each of which I control for in the
following analyses. Perhaps the most important such variables deal directly with the location of SQ.
In particuar, I control for the ideological location of SQ and how extreme SQ is. Some theoretical
28

models (e.g. Dziuda and Loeper, 2018) suggest that political systems with large numbers of veto players
(such as the U.S.) can exhibit over-time policy biases. erefore, I consider whether conservative or
liberal policies appear more or less likely to be met viable or messaging proposals. Second, I control
for the overall extremity of the status quo policy in question, captured by the absolute value of the SQ
Location term. Here, the expectation is that the estimated coefficient is positive: for policies that are to
the far right or far left, it is easier to make a proposal that both AS and Rwould accept. Not only is the
range of possible such offers larger, but the chance is greater that this range will include the proposer’s
ideal point.
In addition to these SQ-related variables, I include a variety of variables related to the proposer
herself. First, I control for majority status. Assuming that majority members face an added incentive
to pass legislation (in order for their party to be viewed as competent), such members may be more
willing to settle for less than their ideal point. Moreover, since it is likely that agenda-setters (themselves
members of the majority party) are more likely to move on legislation authored by their copartisans,
majority members may face a higher potential payoff for introducing viable legislation, all else equal.
Second, I control for a member’s gender. Recent research (Volden, Wiseman and Wittmer, 2013)
has suggested that female members are more effective legislators, which could be related to their propensity
to offer viable legislative proposals. If so, the coefficient on Female would be positive. ird, I control
for party (captured by the binary variable Democrat). Given that previous literature has underscored
the fact that Republicans have polarized to a greater extent than have Democrats (McCarty, Poole and
Rosenthal, 2008), and given that some literature has suggested that Republicans are more ideologically
motivated than are Democrats (Grossman and Hopkins, 2016), Republicans may be more likely to
make messaging proposals which more faithfully reflect their ideal points. Finally, I control for a
sponsor’s ideological extremity. Members on the far reaches of the ideological spectrum not only may
be less interested in compromise, but they also have, mathematically speaking, fewer opportunities
within the policy space for offering viable proposals (assuming they wish to offer policies that improve
upon the status quo for themselves). Consequently, I expect Ideological Extremity to be negatively
associated with viable proposals. In place of these member-level covariates, I also estimate a series of
models with member-specific fixed effects, summarized in Appendix F, which uncover substantively
similar results to those presented below.
29

Empirical Approach and Results
In order to test H1, H2and H3, I run a series of bill-level logistic regression models, designed to predict
whether a member met a targeted status quo policy with a viable proposal. Given that my theory makes
predictions about the locations and timing of bills—and not individual legislators’ propensities to offer
certain kinds of proposals per se—a bill-level model is more appropriate for this analysis than, say, a
member-level model.
In order to test the above hypotheses, I run three series of logistic regressions with cluster-robust
standard errors by Congress. In particular, I regress the binary Yijm (viable proposal) variable on each
of the three indicator variables associated with each hypothesis separately. One possibility would be to
estimate a single model with all three variables. However, given that H2and H3both respond to policy
acceleration, a unified model would disallow separate testing of these two hypotheses. As a result,
it would be impossible to discern whether both hypotheses receive support, or whether one type of
acceleration is driving any observed result. Moreover, because the policy acceleration and deceleration
variables are exhaustive and mutually exclusive in these data, inclusion of all terms in the same model
would force one of the terms into the model’s constant term. is would obfuscate the interpretation
of the model, for either policy deceleration or acceleration. ese challenges notwithstanding, I do
estimate a unified deceleration-acceleration model in Appendix A, the results of which are consistent
with each of the above hypotheses.
Within each of the models presented below, I vary the subsample, fixed effects, and clustering of
standard errors in several different ways, in order to demonstrate the robustness of the results. I provide
additional context about these results in each respective section. However, the overall sample of bills I
include in the models presented here is of particular note. at is, in all models, I remove SQ policies
located within [AS, R] (the static gridlock interval), since such policies cannot—by definition—be met
with a viable proposal. Indeed, no proposal exists that would improve upon such status quo policies for
both the left and right end points of the static gridlock interval.14 Interestingly, though, the percentage
of all proposals found in [AS, R], depicted in Figure 7 is highest within Congresses exposed to policy
deceleration, lending some support to the notion that policy deceleration encourages the introduction
of non-viable proposals.
14Which, of course, is why this interval is considered in equilibrium in static policy change models.
30

Figure 7: Percentage of Proposals Located within [AS, R], By Congress
H1: Policy Deceleration
According to the logic delineated in Scenario 1, AS faces incentives to decelerate policy change when
she anticipates that she will retain control of AS and gain ground with R(i.e., face a more proximate
R), following the upcoming election. In particular, AS has incentives to put off changes to status
quo policies located on the opposite side of Rfrom AS (see Figure 2). Understanding this incentive
structure, I argue, members of Congress will be reticent to engage in viable bill-writing. Inasmuch
as such legislation is costlier to draft than legislation lying closer to the member’s ideal point, when
members do decide to address SQ in this region, they will be unlikely to offer viable proposals. Instead,
they may be better served to propose messaging legislation closer to their ideal point.
Table 1 displays strong evidence in favor of strategic member behavior when facing a deceleration
environment. Indeed, throughout all model specifications, the association between SQ’s location
within the deceleration interval and the introduction of viable proposals is negative and statistically
significant. Model 1 makes the fullest use of the entire available sample of bills located outside [AS,R],
as noted above, and they each display a strong, negative association between the introduction of viable
proposals and a SQ’s location within the deceleration region.
ese results are robust to a wide variety of robustness checks, some of which are also displayed
in Table 1. First and foremost, given that Ris typically located within the Senate, I rerun all models
using only bills originating in the House of Representatives. Implicit in H1is the assumption that bill
sponsors—regardless of chamber—consider the location of R, as they decide how they will respond to
31

Table 1: Policy Deceleration and Viable Proposals (Scenario 1)
Dependent variable:
Introduction of Viable Proposal
(1) (2) (3) (4)
SQ ∈Deceleration Region −0.735∗∗∗ −0.827∗∗ −1.994∗∗∗ −1.485∗∗∗
(0.263) (0.387) (0.358) (0.458)
Majority Status 1.755∗∗∗ 3.466∗∗∗ −1.120 −4.302∗∗
(0.373) (0.778) (0.735) (2.152)
SQ Location −0.059 −0.020 0.005 −0.218
(0.068) (0.102) (0.171) (0.247)
|SQ Location| 1.860∗∗∗ 2.026∗∗∗ 2.437∗∗∗ 2.528∗∗∗
(0.155) (0.264) (0.232) (0.343)
Female −0.390 −0.900∗∗ −0.948∗∗ −0.966∗
(0.280) (0.402) (0.416) (0.573)
Democrat 0.743∗∗∗ 1.878∗∗∗ 2.121∗∗∗ 3.365∗∗∗
(0.285) (0.529) (0.640) (0.938)
Ideological Extremity −1.113∗∗∗ −1.156∗∗ −1.193∗∗∗ −1.720∗∗∗
(0.242) (0.456) (0.374) (0.664)
Constant −4.840∗∗∗ −7.080∗∗∗ −2.726∗∗ 0.073
(0.614) (1.123) (1.151) (2.666)
Observations 753 445 507 320
Sample Full House Opposite AS House/Opp. AS
Log Likelihood −286.574 −147.314 −154.140 −88.533
Akaike Inf. Crit. 589.147 310.628 324.281 193.066
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
32

policy acceleration and deceleration. is assumption is most reasonable for sponsors in the Senate,
given that Ris located within the sponsor’s own chamber. In the House, however, it may be less
reasonable to assume that sponsors bear in mind the preferences of R, when it is located in a different
chamber. us, to demonstrate that the results are not driven solely by sponsorship behavior in the
Senate alone, Models 2 and 4 are estimated with House-only data, with similar results to those found
in Model 1.
In addition to this robustness check, I rerun Models 1 and 2 using only SQ policies located beyond
R on the opposite side of AS. at is, because the deceleration region is located outside of [AS, R] to the
opposite side of AS, restricting the analysis to these regions ensures that control units are drawn from
as similar an ideological region as possible. I restrict the sample in this way in Models 3 and 4, which
provide perhaps the most rigorous test of H1found in Table 1: by holding the SQ region fixed, Models
3 and 4 test specifically whether the theory isolates the proper timing for deceleration, and not just
spatial bill location. As Table 1 indicates, the theory performs quite well. Across both models, policy
deceleration Congresses experience far fewer viable proposals, given their associated status quos. is
time, however, the effect size is much larger than in Models 1 and 2.
33

Table 2: Scenario 1 Results with Fixed Effects
Dependent variable:
Introduction of Viable Proposal
(5) (6) (7) (8) (9) (10) (11) (12)
SQ ∈Decel. Region −0.889∗∗∗ −0.983∗∗ −2.574∗∗∗ −2.339∗∗∗ −0.572∗∗ −0.679∗−2.022∗∗∗ −1.415∗∗∗
(0.302) (0.444) (0.471) (0.629) (0.276) (0.387) (0.378) (0.503)
Majority Status 1.843∗∗∗ 3.600∗∗∗ −1.308 −5.908∗1.965∗∗∗ 4.065∗∗∗ −0.754 −3.623∗∗
(0.448) (0.991) (1.146) (3.355) (0.334) (0.630) (0.614) (1.628)
SQ Location −0.137 −0.095 −0.014 −0.322 −0.095 −0.077 −0.041 −0.298
(0.084) (0.136) (0.276) (0.382) (0.070) (0.104) (0.149) (0.296)
|SQ Location| 2.081∗∗∗ 2.265∗∗∗ 3.320∗∗∗ 3.448∗∗∗ 2.051∗∗∗ 2.412∗∗∗ 3.003∗∗∗ 3.231∗∗∗
(0.206) (0.368) (0.417) (0.622) (0.179) (0.276) (0.323) (0.467)
Female −0.361 −0.845∗−1.063∗∗ −1.381 −0.368 −0.901∗−1.022∗∗ −1.039
(0.306) (0.469) (0.499) (0.904) (0.326) (0.541) (0.521) (0.792)
Democrat 0.882∗∗ 2.005∗∗∗ 2.304∗∗ 3.685∗∗∗ 0.896∗∗∗ 2.518∗∗∗ 2.605∗∗∗ 4.439∗∗∗
(0.352) (0.672) (1.042) (1.356) (0.298) (0.514) (0.605) (1.285)
Ideo. Extremity −0.968∗∗∗ −1.136∗∗ −1.363∗∗ −1.915∗−1.201∗∗∗ −1.455∗∗∗ −1.093∗∗∗ −2.170∗∗∗
(0.297) (0.558) (0.632) (1.117) (0.260) (0.481) (0.368) (0.724)
Constant −6.718∗∗∗ −8.371∗∗∗ −4.642∗−0.431 −5.676∗∗∗ −8.382∗∗∗ −4.721∗∗∗ −1.969
(1.140) (1.797) (2.697) (4.388) (0.688) (1.139) (1.074) (1.867)
Observations 710 419 478 299 753 445 507 320
Effects Issue Issue Issue Issue Comm. Comm. Comm. Comm.
Sample Full House Opp. AS House/Opp. AS Full House Opp. AS House/Opp. AS
Log Likelihood −246.540 −127.722 −118.617 −65.262 −264.986 −127.237 −131.377 −73.006
Akaike Inf. Crit. 547.079 309.443 291.235 184.524 613.973 332.474 346.754 224.013
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
34

e results in Table 1 are similarly robust to the introduction of fixed effects into the base model,
depicted in Table 2, ensuring that unobserved confounds at the issue-, committee-, and (in the appendix)
member-level are not driving the observed results. Fixed effects by a bill’s primary issue area are
drawn from the Congressional Bills Project (Adler and Wilkerson, 2006), which uses classifications
provided by the Policy Agendas Project. Fixed effects by committee of referral are also drawn from the
Congressional Bills Project.15 Here, it is notable that the magnitude of the association between location
in the deceleration interval and probability of offering a viable proposal does vary based on fixed effect
type. However, the results remain substantively and statistically significant, again providing evidence
in favor of H1.
Finally, the basic result captured in Model 1 is robust to tests that account for potential differences
between long-serving and new members of Congress. ough not presented here, I replicate each of
the above analyses and subsequent robustness checks solely using bills authored by members who served
for all five Congresses in the sample, an analysis I summarize in Appendix B. Much as with the other
robustness checks above, the results remain statistically and substantively similar to the results presented
in the tables above.
In sum, while the effects sizes vary somewhat between the models, each of the models exhibits
a substantively meaningful association with the introduction of viable proposals. In Model 1, which
includes the smallest association between deceleration region, status quo policies within the deceleration
region are 15 percent less likely to be met with a viable proposal than similar bills not subject to
deceleration dynamics, holding all other variables at their means or (in binary cases) their optimal
levels. In Model 5, the association is even larger: bills subject to deceleration are 24 percent less likely
than similar bills to be met with a viable proposal. When members face a scenario under which they
should expect policy deceleration, then, they appear to be introduce bills that are systematically less
likely to improve upon the status quo for pivotal actors.
Before discussing policy acceleration dynamics, a few other results merit mention. First, as expected,
majority party members are significantly more likely to offer viable proposals, all else equal—at least
in Models 1 and 2. In Models 3 and 4, however, the opposite trend emerges. Indeed, for policies that
lie opposite AS—and therefore, opposite most or all majority party members—legislators within the
15In addition to these fixed effects, I estimate a separate set of models with errors clustered by issue area and committee
of referral. Results from these models are nearly identical to the primary results presented here, so they are not presented
here.
35

majority are less likely to offer viable proposals. In other words, when Republicans seek to address liberal
status quo policies (and vice versa) as members of the majority, they appear to be attempting to move
policy beyond what Rwill tolerate. is, of course, does not mean that amendments and committee
activity will not moderate the bill, but this trend reversal is notable nonetheless. e pattern holds in
Table 2, when fixed effects are included.
Beyond majority party membership, |SQ Location|, Democrat, and Ideological Extremity all behave
as expected. Democrats do appear systematically more likely to offer viable proposals, consistent with
asymmetries highlighted in previous work. Similarly, more extreme members demonstrate a lower
propensity for offering viable proposals, all else equal. Also, consistent with spatial constraints associated
with moderate versus extreme status quo policies, moderate status quo policies are less likely to be paired
with a viable proposal (and extreme status quos more likely). Interestingly, however, female members
of Congress do not appear to offer more viable proposals than do their male counterparts. While the
effect is not consistently significant, female members are if anything less likely to offer such proposals.
Taken together, the evidence presented in Table 1 is consistent with theoretical expectations. Under
conditions of policy deceleration, members of Congress appear less likely to offer viable proposals for
consideration.
H2: Policy Acceleration when AS Does Not Change
As detailed above, when AS is not expected to change but Ris expected to move away from AS,AS
faces an incentive to accelerate policymaking for a particular set of SQ. More specifically, AS should
focus her policymaking energy on SQ > R +|SQ −R|,16 as movements of those SQ will become
either impossible or smaller following the upcoming election. Understanding this dynamic, members
of Congress will, I argue, propose legislation that is viable, rather than messaging, within this region.
at is, they will generate proposals—for SQ within the acceleration region specifically—that would
pass through Congress if given the opportunity, since the overall probability of those policies receiving
agenda space should be higher.
Table 3 investigates whether or not SQiwithin the acceleration region are in fact met with viable
proposals at a higher rate than those outside of that region. As in Table 1, Models 1 and 2 make full
use of all data outside the static gridlock region, with Model 2 focusing on House-only proposals and
16When Democrats control AS; the analogous region for Republican control is SQ < R − |SQ −R|.
36

Model 1 making use of all proposals. As Table 3 depicts, support for H2is mixed. In particular, while
the results are positive and significant in Model 1 (as well as models in the robustness checks that also
pool over both chambers), results in Model 2 (and other models restricting the sample to the House
bills) are not significant. ese positive results are consistent with H2, although the weak House-only
results indicate that incentives for viable proposal-making are not as strong as in the other cases.
One reason for this relative weakness might be due to the uncertainty associated with the electoral
outcomes of the treated Congress under this scenario, the 114th Congress. Under this scenario, the
IEM predicted for most of 2015 and 2016 that Republicans would incur losses in the Senate, due in
part to the unpopularity of their presidential candidates. Such a dynamic should have led Republicans
to push policymaking forward; however, many Republicans may have reasonably believed that the
threat of losing the Senate was not as great as some worried.17 us, given the uncertainty of 2016
electoral prospects, the support for H2shown here is perhaps understandably weaker than in the other
scenarios. Nevertheless, among models exhibiting a statistically significant association between viable
proposal-making and SQ location in the acceleration region, acceleration-region bills are between 7
and 11 percent more likely to be met with viable proposals than are similar bills outside this region.
ese results are similar in the first set of robustness checks, found in Models 3 and 4. As in
Scenario 1, these models focus only on SQ lying opposite Rfrom AS. Here again, Models 3 and
4 offer the most difficult test of the theory, as they compare SQ that are similar in spatial location and
other factors but differ in their exposure to electoral incentives for acceleration. Like with Models 1
and 2, the results are positive accross both models, with only the pooled-chamber results (Model 3)
achieving statistical significance. Similarly, the results do not change appreciably when fixed effects are
incorporated, reported in Table 4. at is, an SQ’s location within the Scenario 2 acceleration region
is consistently positively associated with the introduction of a viable proposals in all but a handful of
the House-only models.
17In fact, Republicans surprisingly did not incur losses in 2016 and even won the White House.
37

Table 3: Policy Acceleration and Viable Proposals (Scenario 2)
Dependent variable:
Introduction of Viable Proposal
(1) (2) (3) (4)
SQ ∈Acceleration Region 0.700∗∗ 0.590 1.225∗∗∗ 0.623
(0.333) (0.475) (0.402) (0.574)
Majority Status 1.691∗∗∗ 3.380∗∗∗ −0.640 −4.567
(0.368) (0.763) (0.581) (3.325)
SQ Location 0.032 0.068 0.155 −0.268
(0.063) (0.093) (0.138) (0.258)
|SQ Location| 1.869∗∗∗ 2.026∗∗∗ 2.262∗∗∗ 2.340∗∗∗
(0.155) (0.259) (0.226) (0.302)
Female −0.391 −0.893∗∗ −0.800∗∗ −0.893
(0.281) (0.412) (0.376) (0.548)
Democrat 0.877∗∗∗ 1.984∗∗∗ 2.205∗∗∗ 4.015∗∗∗
(0.298) (0.521) (0.554) (1.019)
Ideological Extremity −1.098∗∗∗ −1.182∗∗∗ −1.030∗∗∗ −1.562∗∗
(0.237) (0.440) (0.335) (0.612)
Constant −5.283∗∗∗ −7.432∗∗∗ −4.237∗∗∗ −0.621
(0.628) (1.165) (0.890) (3.605)
Observations 753 445 507 320
Sample Full House Opposite AS House/Opp. AS
Log Likelihood −288.616 −149.441 −170.486 −94.761
Akaike Inf. Crit. 593.231 314.882 356.973 205.522
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
38

Table 4: Scenario 2 Results with Fixed Effects
Dependent variable:
Introduction of Viable Proposal
(1) (2) (3) (4) (5) (6) (7) (8)
SQ ∈Accel. Region 0.890∗∗ 0.835 1.510∗∗∗ 1.255∗0.644∗0.590 1.273∗∗∗ 0.595
(0.385) (0.552) (0.477) (0.691) (0.367) (0.515) (0.448) (0.633)
Majority Status 1.770∗∗∗ 3.534∗∗∗ −0.728 −6.385 1.923∗∗∗ 4.015∗∗∗ −0.256 −3.297∗
(0.439) (0.998) (0.787) (5.840) (0.336) (0.632) (0.564) (1.760)
SQ Location −0.025 0.018 0.177 −0.380 −0.019 0.005 0.107 −0.336
(0.074) (0.118) (0.189) (0.374) (0.071) (0.106) (0.144) (0.313)
|SQ Location| 2.079∗∗∗ 2.271∗∗∗ 2.856∗∗∗ 2.981∗∗∗ 2.065∗∗∗ 2.415∗∗∗ 2.848∗∗∗ 3.068∗∗∗
(0.205) (0.370) (0.366) (0.524) (0.179) (0.275) (0.306) (0.442)
Female −0.368 −0.850∗−0.812∗−1.162 −0.373 −0.881 −0.888∗−1.092
(0.308) (0.481) (0.439) (0.805) (0.326) (0.537) (0.483) (0.755)
Democrat 1.071∗∗∗ 2.220∗∗∗ 2.431∗∗∗ 4.813∗∗∗ 1.027∗∗∗ 2.621∗∗∗ 2.708∗∗∗ 5.057∗∗∗
(0.366) (0.678) (0.767) (1.475) (0.314) (0.522) (0.590) (1.358)
Ideological Extremity −0.973∗∗∗ −1.227∗∗ −1.152∗∗ −1.686∗−1.193∗∗∗ −1.522∗∗∗ −1.025∗∗∗ −2.193∗∗∗
(0.298) (0.539) (0.494) (0.907) (0.259) (0.480) (0.341) (0.723)
Constant −7.138∗∗∗ −8.782∗∗∗ −5.874∗∗∗ −1.022 −6.077∗∗∗ −8.651∗∗∗ −6.241∗∗∗ −2.987
(1.162) (1.882) (1.864) (6.260) (0.690) (1.160) (0.996) (1.839)
Observations 710 419 478 299 753 445 507 320
Effects Issue Issue Issue Issue Comm. Comm. Comm. Comm.
Sample Full House Opposite AS House/Opp. AS Full House Opp. AS House/Opp. AS
Log Likelihood −248.982 −129.846 −137.978 −74.182 −265.656 −128.165 −144.205 −76.869
Akaike Inf. Crit. 551.965 313.691 329.957 202.364 615.311 334.330 372.410 231.738
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
39

Given that the models presented here are highly similar to those examined in the deceleration
analysis, most of the additional covariates behave similarly in Table 2 as in Table 1. However, it
is worth noting that, for Models 5 and 6, Majority Status falls out of statistical significance in the
opposite-AS models. Apart from this difference, though, much remains similar. Here again, sponsor
characteristics such as Ideological Extremity and party (Democrat) are associated with a lower and higher
probability for viable proposals, respectively. SQ extremity is also again positively associated with viable
proposal-making.
Taken together, these models provide some support for the hypothesis that agenda-setting dynamics
generated by electoral expectations—this time providing incentives for acceleration—influence members’
propensities for offering viable or messaging legislation.
H3: Policy Acceleration when AS Changes
Whereas AS is expected to remain under control of the same party in Scenarios 1 and 2, it is expected
to change in Scenario 3. is anticipated change generates significant discrepancies in expected policy
outcomes in Periods 1 and 2, across large portions of the policy space. Indeed, as captured in Proposition
3, AS faces an incentive to accelerate policymaking for all SQ lying outside the static gridlock interval.
Consequently, for status quo policies located outside the static gridlock interval, members of Congress
should be more likely to offer viable proposals when they expect AS to change parties—as described
in H3.
Tests of H3reveal strong support for the idea that members of Congress respond to possible changes
in AS by proposing viable legislation. Table 5 summarizes these results. As in the previous two analyses,
Model 1 makes full use of the dataset, while Model 2 focuses on House bills alone. As illustrated in
Table 5, H3receives strong support across each of these models. Indeed, when members target SQ
inside the acceleration region, they appear far more likely to offer a viable proposal than when they
target otherwise similar bills not located within this region. is association remains strong regardless
of whether or not the models pool across a bill’s chamber of origin. Here again, the results are robust to
the inclusion of fixed effects: Models 3 and 4 include issue-area fixed effects while Models 5-6 include
committee fixed effects, with each model displaying the same basic (negative) result.
Holding all other variables at their means or optimal values, bills with SQ located in Scenario 3’s
acceleration region are viable with probability between 0.91 or 0.93 (depending upon the specific model
40

Table 5: Policy Acceleration and Viable Proposals (Scenario 3)
Dependent variable:
Introduction of Viable Proposal
(1) (2) (3) (4) (5) (6)
SQ ∈Acceleration Region 2.789∗∗∗ 2.450∗∗∗ 3.196∗∗∗ 3.061∗∗∗ 2.949∗∗∗ 2.353∗∗∗
(0.410) (0.622) (0.466) (0.874) (0.372) (0.543)
Majority Status 1.234∗∗∗ 2.389∗∗∗ 1.336∗∗∗ 2.195∗1.482∗∗∗ 2.962∗∗∗
(0.421) (0.890) (0.513) (1.240) (0.359) (0.653)
SQ Location −0.023 0.011 −0.101 −0.072 −0.072 −0.060
(0.078) (0.107) (0.099) (0.149) (0.074) (0.104)
|SQ Location| 2.207∗∗∗ 2.217∗∗∗ 2.582∗∗∗ 2.551∗∗∗ 2.423∗∗∗ 2.618∗∗∗
(0.174) (0.272) (0.253) (0.383) (0.206) (0.298)
Female −0.340 −1.029∗∗ −0.291 −1.120∗∗ −0.326 −1.020∗
(0.320) (0.481) (0.354) (0.539) (0.348) (0.573)
Democrat −0.308 0.580 −0.342 0.314 −0.228 1.245∗∗
(0.385) (0.691) (0.490) (0.958) (0.353) (0.585)
Ideological Extremity −1.141∗∗∗ −1.110∗∗ −0.978∗∗∗ −0.897 −1.253∗∗∗ −1.356∗∗∗
(0.281) (0.480) (0.326) (0.590) (0.292) (0.489)
Constant −5.407∗∗∗ −6.896∗∗∗ −8.118∗∗∗ −8.961∗∗∗ −6.191∗∗∗ −8.158∗∗∗
(0.671) (1.228) (1.360) (1.898) (0.755) (1.152)
Observations 753 445 710 419 753 445
Effects None None Issue Issue Committee Committee
Sample Full House Full House Full House
Log Likelihood −250.768 −136.182 −208.411 −114.561 −228.792 −118.775
Akaike Inf. Crit. 517.535 288.363 470.821 283.122 541.584 315.549
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
41

referenced). Similar bills located outside the acceleration region are met with a viable proposal with a
much smaller probability, between 0.40 and 0.53. is means that electoral incentives for acceleration
are associated with an impressive 48 - 51 percentage increase in viable proposal-making instead of
messaging proposal-making.
Much as in previous models, Ideological Extremity of the proposer and the extremity of the SQ are
negatively associated with viable proposal-making. Moreover, Majority status again remains a positive
and significant predictor of the introduction of viable proposals. Sponsor gender, on the other hand,
is inconclusively associated with viable proposal-making—though again, if anything, female members
are less likely to introduce viable proposals. Finally, there does not appear to be an association between
party and viable or messagine proposal-making in these models—a result that differs from the two
previous analyses.
Taken together, the results from this third and final analysis are strongly consistent with H3. When
control of AS is expected to change from one party to the other, AS faces strong incentives to accelerate
policymaking for SQ lying outside the static gridlock interval. Understanding this dynamic, members
of Congress appear more likely to offer viable proposals for consideration, rather than messaging bills.
Discussion and Conclusion
For several decades now, Washington journalists and scholars of Congress have derided the apparent
lack of seriousness with which members engage with the lawmaking process, underscoring how policymaking
activity only occasionally exhibits actual potential for altering the status quo. Still, in spite of these
observations, academic research has generally neglected to explain when and why members offer viable
or messaging legislation—or how electoral expectations may influence their legislative activities more
broadly. In this study, I provide a theoretical framework for understanding how fluctuations in competition
over partisan control of pivotal actors in Congress influence members’ bill sponsorship activity. Empirical
examinations of this framework demonstrate that expectations over future partisan advantages and
disadvantages seem to influence the types of legislation members are willing to sponsor.
More than influencing observed sponsorship patterns, these results speak to how Congress’s most
fundamental institutional features influence a core responsibility of individual legislators. Indeed,
given Congress’s frequent elections, the possibility for changes in pivotal actors’ preferences is inherent
to institutional designs established in Article I of the Constituion. Moreover, for over 100 years,
42

party leaders have wielded strong agenda control within the House of Representatives (Gailmard and
Jenkins, 2007). When combined with highly insecure majority control, these factors generate incentive
structures that alter members’ calculations about how and when they should engage with the policymaking
process. Consequently, while members do ultimately decide whether and how to draft legislation, this
study demonstrates that such lawmaking decisions derive not simply from members’ personal style or
skill, but also from contextual factors that can coax members into more or less productive patterns of
legislative activity.
e influence of electoral and institutional context on individual behavior, of course, complicates
the measurement of member-level attributes, such as effectiveness (Volden and Wiseman, 2014). Not
only may members’ revealed legislative effectiveness derive from more than their individual skills and
best practices, but the context that influences their application of those skills changes from Congress
to Congress. us, comparing a member’s observed effectiveness across Congresses may prove more
difficult than previously suggested. Further still, given that moderate members of Congress are, by
assumption, more comfortable than other members to offer viable legislation (since policy movements
to their ideal point are more likely to be acceptable to pivotal actors in Congress), some members may
enjoy inherent advantages in how effective they appear to outside observers. is is not to say that
outcomes-based measures of effectiveness cannot be useful, nor that legislative best practices cannot be
gleaned from members who score highly in these metrics. Rather, future research should build upon
these measures by incorporating more context-invariant information, such as data on policy valence
(c.f., Hitt, Volden and Wiseman 2017), into measures of legislator effectiveness.
More broadly, given that Congress’s electoral history has varied considerably in terms of competitiveness,
understanding how electoral dynamics influence legislative behavior remains a crucial topic for future
research. Indeed, while Congress has experienced prolonged eras of partisan dominance (e.g., Democratic
rule for much of the 20th Century), intense competition over control of congressional majorities has
developed since the 1980s and 90s, fundamentally altering how members of Congress approach their
work as legislators (Lee, 2016). In spite of these dramatic differences over time, current models of
policy change and bills sponsorship activity rarely consider how electoral competition might influence
the strategic environment within which lawmakers propose laws and policy change occurs. is study
provides a framework for understanding how these electoral dynamics influence not only when bills
should pass into law, but also how members of Congress may respond to the differential incentives
43

introduced by various electoral regimes.
44

References
Adler, E Scott and John Wilkerson. 2006. “Congressional bills project.” NSF
880066(00880061):213–227.
Buisseret, Peter and Dan Bernhardt. 2017. “Dynamics of policymaking: Stepping back to leap forward,
stepping forward to keep back.” American Journal of Political Science 61(4):820–835.
Clinton, Joshua D. 2017. “Coding the Ideological Direction and Content of Policies.” Annual Review
of Political Science 20(1):433–450.
Clinton, Joshua, Simon Jackman and Douglas Rivers. 2004. “e statistical analysis of roll call data.”
American Political Science Review 98(2):355–370.
Cox, Gary W and Mathew D McCubbins. 2005. Setting the agenda: Responsible party government in
the US House of Representatives. Cambridge University Press.
Cox, Gary W and William C Terry. 2008. “Legislative productivity in the 93d-105th congresses.”
Legislative Studies Quarterly 33(4):603–618.
Crosson, Jesse M. 2019. “Stalemate in the States: Agenda Control Rules and Policy Output in American
Legislatures.” Legislative Studies Quarterly 44(1):3–33.
Crosson, Jesse M., Alexander C. Furnas and Geoffrey M. Lorenz. n.d. “Estimating Bill Proposal Point
Estimates Using Position-Taking Data.”.
Dziuda, Wioletta and Antoine Loeper. 2018. “Dynamic pivotal politics.” American Political Science
Review pp. 1–22.
Gailmard, Sean and Jeffrey A. Jenkins. 2007. “Negative agenda control in the senate and house:
Fingerprints of majority power.” The Journal of Politics 69(3):689–700.
Garand, James C and Kelly M Burke. 2006. “Legislative activity and the 1994 republican takeover:
Exploring changing patterns of sponsorship and cosponsorship in the US House.” American Politics
Research 34(2):159–188.
Gelman, Jeremy. 2017. “Rewarding Dysfunction: Interest Groups and Intended Legislative Failure.”
Legislative Studies Quarterly 42(4):661–692.
Grossman, Matt and DA Hopkins. 2016. Asymmetric politics. Oxford, United Kingdom: Oxford
University Press.
Hall, Richard L. 1996. Participation in congress. Yale Univ Pr.
Hitt, Matthew P, Craig Volden and Alan E Wiseman. 2017. “Spatial models of legislative effectiveness.”
American Journal of Political Science 61(3):575–590.
Krehbiel, Keith. 1998. Pivotal politics: A theory of US lawmaking. University of Chicago Press.
45

Lee, Frances E. 2016. Insecure majorities: Congress and the perpetual campaign. University of Chicago
Press.
Mayhew, David R. 1974. Congress: The electoral connection. Vol. 26 Yale University Press.
McCarty, Nolan, Keith T Poole and Howard Rosenthal. 2008. Polarized America: The dance of ideology
and unequal riches. MIT Press.
Peress, Michael. 2013. “Estimating proposal and status quo locations using voting and cosponsorship
data.” The Journal of Politics 75(3):613–631.
Poole, Keith T and Howard Rosenthal. 1997. Congress: A political-economic history of roll call voting.
Oxford University Press.
Rocca, Michael S and Stacy B Gordon. 2010. “e position-taking value of bill sponsorship in
congress.” Political Research Quarterly 63(2):387–397.
Schiller, Wendy J. 1995. “Senators as political entrepreneurs: using bill sponsorship to shape legislative
agendas.” American Journal of Political Science February:186–203.
Sulkin, Tracy. 2005. Issue politics in Congress. Cambridge University Press.
ieme, Sebastian. 2018. “A Direct Test of Committee Gatekeeping.” Available at SSRN 3098587 .
Volden, Craig and Alan E Wiseman. 2014. Legislative effectiveness in the United States congress: The
lawmakers. Cambridge University Press.
Volden, Craig, Alan E Wiseman and Dana E Wittmer. 2013. “When are women more effective
lawmakers than men?” American Journal of Political Science 57(2):326–341.
Wawro, Gregory. 2000. Legislative entrepreneurship in the US House of Representatives. University of
Michigan Press.
Woon, Jonathan and Ian Palmer Cook. 2015. “Competing gridlock models and status quo policies.”
Political Analysis 23(3):385–399.
46

Appendix: Progress or Principle
A: Unified Acceleration Models
While the primary analysis in the paper separately examines policy acceleration under H2and H3, the
results are robust to inclusion of all terms in a single model. Indeed, when all policy-acceleration-related
variables are included in a single model, conditions for policy acceleration are consistently positively
associated with the introduction of viable proposals. Here again, because all Congresses in the sample
are subject to either policy acceleration or deceleration, inclusion of a unified policy acceleration term
forces policy deceleration into the constant of the model. Even still, the constant term points in the
expected (negative) direction.
Table A1 summarizes these results. In the table, odd-numbered models make use of all available
data (less bills included in [AS,R]), while even-numbered models make use of bills first introduced in
the House—analogous to the presentation of results in the main text. Models 1 and 2 include no fixed
effects, Models 3 and 4 include major issue topic fixed effects, and Models 5-6 include fixed effects
for primary committee of referral. Finally, Models 7 and 8 winnow the sample to bills introduced
by members who served in all five Congresses in the sample. In each model, errors are clustered by
Congress.
1

Table 1: Unified Acceleration Models (H2 and H3)
Dependent variable:
Introduction of Viable Proposal
(1) (2) (3) (4) (5) (6) (7) (8)
Policy Acceleration 2.117∗∗∗ 1.618∗∗∗ 2.493∗∗∗ 2.039∗∗∗ 2.233∗∗∗ 1.618∗∗∗ 2.058∗∗∗ 1.382∗∗
(0.274) (0.354) (0.328) (0.409) (0.299) (0.354) (0.370) (0.572)
Majority Status 1.224∗∗∗ 2.803∗∗∗ 1.275∗∗ 2.727∗∗ 1.454∗∗∗ 2.803∗∗∗ 1.195∗∗ 3.624∗∗∗
(0.404) (0.798) (0.501) (1.075) (0.357) (0.798) (0.518) (1.056)
SQ Location 0.108 0.112 0.027 0.038 0.061 0.112 0.029 −0.017
(0.074) (0.102) (0.091) (0.138) (0.073) (0.102) (0.097) (0.154)
|SQ Location| 2.130∗∗∗ 2.171∗∗∗ 2.522∗∗∗ 2.555∗∗∗ 2.358∗∗∗ 2.171∗∗∗ 2.065∗∗∗ 2.485∗∗∗
(0.170) (0.271) (0.251) (0.396) (0.201) (0.271) (0.214) (0.376)
Female −0.398 −0.935∗∗ −0.363 −0.963∗−0.339 −0.935∗∗ −0.204 −1.572∗∗
(0.314) (0.453) (0.356) (0.537) (0.344) (0.453) (0.396) (0.664)
Democrat 0.462 1.383∗∗∗ 0.604 1.390∗∗ 0.700∗∗ 1.383∗∗∗ 0.714∗2.703∗∗∗
(0.317) (0.525) (0.395) (0.675) (0.313) (0.525) (0.400) (0.792)
Ideological Extremity −1.270∗∗∗ −1.294∗∗∗ −1.250∗∗∗ −1.290∗∗ −1.362∗∗∗ −1.294∗∗∗ −0.906∗∗ −0.287
(0.260) (0.446) (0.326) (0.558) (0.283) (0.446) (0.394) (0.878)
Constant −5.911∗∗∗ −7.431∗∗∗ −8.018∗∗∗ −9.069∗∗∗ −6.764∗∗∗ −7.431∗∗∗ −6.130∗∗∗ −10.390∗∗∗
(0.686) (1.160) (1.379) (2.086) (0.751) (1.160) (0.891) (1.704)
Observations 753 445 710 419 753 445 409 225
Effects None None Major Topic Major Topic Committee Committee None None
Sample Full House Only Full House Only Full House Only 5-term Members 5-term Members
Log Likelihood −254.075 −138.824 −211.030 −117.081 −233.456 −138.824 −145.398 −58.890
Akaike Inf. Crit. 526.150 293.647 476.059 288.163 550.913 293.647 306.797 133.781
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
2

e analysis presented in the body of the paper presents three sets of models, such that each hypothesis
may be examined individually. However, the results in the paper are robust to the inclusion of both
acceleration and deceleration variables in the same model. Indeed, as shown in Table A1, a unified
policy acceleration term is strongly positively associated with the introduction of viable proposals, across
a wide variety of model specifications. Here again, though, because all Congresses in the sample are
subject to either policy acceleration or deceleration, the deceleration term is forced into the constant
term. Even still, the constant exhibits the expected (negative) sign.
Table A1 demonstrates that these results are robust to a wide variety of model specifications. Models
1 and 2 include no effects, while Models 3 and 4 introduce issue area fixed effects and Models 5-6
introduce committee fixed effects. Each model exhibits a strong, positive relationship between exposure
to conditions for policy acceleration and members’ propensity to introduce viable legislation. Models
7 and 8 confine the sample to bills introduced by members who served in all five Congresses in the
sample. Despite the sample restriction, the results remain substantively and statistically significant.
B: Bill Sponsorships by Members Serving in All Five Congresses
One possible confound for the observed patterns of viable and messaging legislation is that new members
in a given Congress could be artificially deflating the number of viable proposals, due to their relative
lack of legislative effectiveness. at is, rather than being attuned to the strategic dynamics generated by
differences in agenda-setting behavior, newer members simply lack the information and skill necessary
to draft legislation that could pass into law, if brough up for a vote. To address this possiblity, I
re-estimate each of the paper’s models using only bills introduced by members who served in all five
Congresses in my sample.
Table A2 summarizes these results for tests of H1(policy deceleration). Even when restricting the
sample to bills introduced by the aforementioned five-term members of Congress, the results remain
largely robust, with seven of the models exhibit main results that are statistically signifiant. Models
1-4 include no effects, while Models 5-8 introduce issue area fixed effects and Models 9-12 introduce
committee fixed effects. Additionally, as in the main text, odd models make use of the full sample,
while even models restrict the sample to House bills alone. Moreoever, Models 3-4, 7-8, and 11-12,
further restrict the sample to include only SQilying opposite the static gridlock interval relative to AS.
e strongest results are found in these opposite-only models, perhaps due to the fact that they provide
3

the most realistic comparison points for treated bills of interest.
4

Table 2: Policy Deceleration Among Long-Serving Members
Dependent variable:
Introduction of Viable Proposal
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
SQ ∈Decel. −0.574 −1.008 −2.279∗∗∗ −1.672∗∗ −1.051∗∗ −2.029∗∗ −3.399∗∗∗ −4.242∗∗ −0.481 −0.663 −2.988∗∗∗ −1.978
Region (0.362) (0.648) (0.515) (0.709) (0.484) (0.880) (0.866) (1.831) (0.401) (0.695) (0.738) (1.216)
Majority 1.702∗∗∗ 4.186∗∗∗ −1.223 −18.013∗∗∗ 1.849∗∗∗ 5.196∗∗∗ −1.229 −21.840 1.915∗∗∗ 4.688∗∗∗ −1.036 −21.489
(0.458) (1.044) (0.933) (2.340) (0.606) (1.153) (1.793) (3,588.091) (0.437) (1.009) (1.070) (4,074.615)
SQ Location −0.132 −0.138 −0.168 0.145 −0.343∗∗∗ −0.724∗∗ −0.402 −0.643 −0.221∗∗ −0.208 −0.493∗−0.393
(0.087) (0.161) (0.255) (0.499) (0.127) (0.292) (0.545) (1.043) (0.099) (0.183) (0.275) (0.821)
|SQ Location| 1.807∗∗∗ 2.398∗∗∗ 2.601∗∗∗ 2.850∗∗∗ 2.309∗∗∗ 3.655∗∗∗ 4.194∗∗∗ 5.729∗∗∗ 2.046∗∗∗ 2.826∗∗∗ 3.669∗∗∗ 4.397∗∗∗
(0.188) (0.367) (0.359) (0.547) (0.319) (0.690) (0.821) (1.751) (0.237) (0.459) (0.579) (0.993)
Female −0.221 −1.599∗∗ −0.395 −1.459∗−0.267 −1.679∗−0.977 −1.604 −0.152 −1.521∗0.185 −0.866
(0.363) (0.638) (0.473) (0.855) (0.461) (0.887) (0.864) (1.813) (0.431) (0.894) (0.880) (1.735)
Democrat 0.821∗∗ 3.076∗∗∗ 3.221∗∗∗ 3.001 1.141∗∗ 4.280∗∗∗ 4.034∗∗ 4.032 1.051∗∗ 4.089∗∗∗ 4.982∗∗∗ 7.856∗∗
(0.363) (0.879) (0.998) (2.006) (0.479) (1.017) (2.051) (3.280) (0.410) (0.965) (1.174) (3.854)
Ideological −0.582 −0.139 −0.703 −1.370 −0.567 −0.327 −1.013 −2.410 −0.617 −0.587 −0.718 −4.498∗
Extremity (0.369) (1.066) (0.491) (0.930) (0.497) (0.962) (0.844) (1.604) (0.408) (1.057) (0.687) (2.726)
Constant −5.368∗∗∗ −10.079∗∗∗ −4.140∗∗∗ 13.198∗∗∗ −9.831∗∗∗ −20.728∗∗∗ −11.609∗∗∗ 1.118 −6.341∗∗∗ −11.183∗∗∗ −7.694∗∗∗ 14.718
(0.817) (1.555) (1.515) (3.537) (1.819) (7.778) (4.156) (3,588.314) (1.040) (2.283) (2.218) (4,074.620)
Obs. 409 225 260 152 384 212 247 143 409 225 260 152
Effects None None None None Issue Issue Issue Issue Comm. Comm. Comm. Comm.
Chamber Both House Both House Both House Both House Both House Both House
Sample All All Opp. AS Opp. AS All All Opp. AS Opp. AS All All Opp. AS Opp. AS
Log Like. −162.624 −60.806 −74.257 −34.337 −128.713 −42.167 −48.913 −18.940 −146.028 −47.775 −51.786 −18.354
AIC 341.248 137.612 164.515 84.673 311.426 138.334 151.826 91.880 372.056 169.549 183.572 110.708
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
5

Table 3: Policy Acceleration (H2) Among Long-Serving Members
Dependent variable:
Introduction of Viable Proposal
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
SQ ∈Accel. 0.641 0.926 2.120∗∗∗ 1.059 1.051∗2.436∗∗ 3.102∗∗∗ 2.905∗0.633 0.435 Region 0.435 2.488∗∗∗
(0.482) (0.962) (0.684) (1.077) (0.558) (1.109) (1.084) (1.731) (0.534) (1.005) (1.005) (0.924)
Majority 1.642∗∗∗ 4.119∗∗∗ −0.917 −16.693∗∗∗ 1.785∗∗∗ 5.208∗∗∗ −0.986 −16.366 1.860∗∗∗ 4.659∗∗∗ 4.659∗∗∗ −0.483
(0.460) (1.013) (0.877) (2.276) (0.593) (1.153) (1.541) (2,280.062) (0.442) (1.021) (1.021) (1.074)
SQ Location −0.053 −0.006 0.033 0.029 −0.193∗−0.377 −0.050 −0.631 −0.151 −0.132 −0.132 −0.194
(0.080) (0.150) (0.223) (0.465) (0.105) (0.244) (0.399) (0.916) (0.100) (0.183) (0.183) (0.258)
|SQ Location| 1.819∗∗∗ 2.388∗∗∗ 2.427∗∗∗ 2.677∗∗∗ 2.292∗∗∗ 3.687∗∗∗ 3.637∗∗∗ 4.618∗∗∗ 2.057∗∗∗ 2.829∗∗∗ 2.829∗∗∗ 3.441∗∗∗
(0.189) (0.347) (0.341) (0.486) (0.313) (0.680) (0.650) (1.238) (0.237) (0.455) (0.455) (0.550)
Female −0.248 −1.736∗∗∗ −0.358 −1.566∗−0.281 −2.202∗∗ −0.586 −2.252∗−0.176 −1.661∗−1.661∗0.146
(0.370) (0.673) (0.490) (0.844) (0.463) (0.886) (0.726) (1.325) (0.433) (0.883) (0.883) (0.783)
Democrat 0.937∗∗ 3.282∗∗∗ 3.606∗∗∗ 4.268∗∗ 1.295∗∗ 4.909∗∗∗ 4.604∗∗∗ 6.951∗∗ 1.178∗∗∗ 4.203∗∗∗ 4.203∗∗∗ 5.097∗∗∗
(0.402) (0.800) (0.913) (2.041) (0.515) (1.107) (1.745) (3.395) (0.439) (1.012) (1.012) (1.210)
Ideological −0.534 −0.220 −0.421 −1.399 −0.500 −0.705 −0.436 −2.250 −0.582 −0.701 −0.701 −0.408
Extremity (0.355) (0.951) (0.570) (0.924) (0.471) (0.998) (0.937) (1.403) (0.402) (1.062) (1.062) (0.624)
Constant −5.784∗∗∗ −10.472∗∗∗ −6.124∗∗∗ 10.721∗∗∗ −10.387∗∗∗ −22.573∗∗ −13.081∗∗∗ −3.208 −6.737∗∗∗ −11.313∗∗∗ −11.313∗∗∗ −10.265∗∗∗
(0.828) (1.564) (1.529) (3.475) (1.885) (11.211) (3.653) (2,280.139) (1.018) (2.249) (2.249) (2.139)
Obs. 409 225 260 152 384 212 247 143 409 225 225 260
Log Like. −162.987 −61.685 −80.794 −36.790 −130.108 −42.734 −57.138 −22.250 −146.063 −48.151 −48.151 −58.635
AIC 341.974 139.369 177.587 89.580 314.215 139.467 168.277 98.500 372.127 170.301 170.301 197.269
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
6

Table 4: Policy Acceleration (H3) among Long-Serving Members
Dependent variable:
Introduction of Viable Proposal
(1) (2) (3) (4) (5) (6)
SQ ∈Accel. 2.460∗∗∗ 1.457∗∗ 3.204∗∗∗ 1.835∗∗ 2.838∗∗∗ 1.254
Region (0.496) (0.695) (0.713) (0.883) (0.489) (0.795)
Majority Status 1.363∗∗ 3.562∗∗∗ 1.336∗3.865∗∗∗ 1.589∗∗∗ 3.973∗∗∗
(0.537) (1.078) (0.686) (1.180) (0.479) (1.077)
SQ Location −0.100 −0.095 −0.354∗∗ −0.473∗∗ −0.223∗∗ −0.176
(0.100) (0.149) (0.152) (0.235) (0.104) (0.172)
|SQ Location| 2.086∗∗∗ 2.457∗∗∗ 2.860∗∗∗ 3.450∗∗∗ 2.384∗∗∗ 2.886∗∗∗
(0.224) (0.386) (0.456) (0.633) (0.277) (0.463)
Female −0.096 −1.516∗∗ −0.194 −1.764∗0.068 −1.393
(0.384) (0.657) (0.517) (0.910) (0.466) (0.901)
Democrat −0.017 2.302∗∗∗ 0.134 2.992∗∗∗ 0.137 3.230∗∗∗
(0.463) (0.870) (0.617) (1.069) (0.457) (1.070)
Ideological Extremity −0.787∗∗ −0.189 −0.792∗−0.464 −0.778∗−0.501
(0.376) (0.875) (0.446) (0.936) (0.432) (1.039)
Constant −5.701∗∗∗ −9.972∗∗∗ −11.343∗∗∗ −19.588∗∗ −6.861∗∗∗ −10.929∗∗∗
(0.838) (1.619) (2.174) (9.110) (1.124) (2.282)
Observations 409 225 384 212 409 225
Log Likelihood −144.790 −59.602 −109.668 −42.932 −126.715 −47.017
Akaike Inf. Crit. 305.581 135.204 273.337 139.864 333.430 168.034
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
C: Introduction of Additional Control Variables
e models presented in the main text each include a set of control covariates that may influence the
introduction of viable proposals, most of which behave as expected. However, I considered a variety
of other potential confounds, which I present in the tables below. ese include a sponsor’s status as a
committee chair or ranking member, subcommittee chair or ranking member, and member seniority.
By and large these variables are not significantly associated with the introduction of viable or messaging
proposals and were therefore not included in the primary analysis.
Tables A5, A6, and A7 summarize the results of regressions including these additional variables. A5
7

focuses on conditions for policy deceleration (Scenario 1), while A6 and A7 deal with policy acceleration
(Scenarios 2 and 3). Each table considers the same combinations of full-sample and House-only-sample
examinations as presented in the paper’s main analysis. ese specifications do not appear to alter the
models’ primary findings in any notable fashion.
D: Iowa Electronic Market Prices and Extrapolations
Table A8 presents the models used to extrapolate the betting price data generated by the Iowa Electronic
Markets. e extrapolations both extend the data backward in time (generating monthly probabilities
for partisan control by chamber, from 1940 forward) and between individual election markets (i.e.,
the months after one election ends and before the next election market opens for betting). e models
regress monthly price averages for each relevant IEM (along with actual partisan control outcomes,
in order to better anchor the historical predictions) on a variety of covariates that may influence a
politician’s assessment of each party’s chances to capture the majority in a chamber. Both models are
logistic regressions estimated via Maximum Likelihood, and they were selected based on an iterative
process that compared predicted electoral probabilities with those actually observed in the IEM data.
Various automated model selection techniques, such as LASSOPlus (Ratkovic and Tingley 2018), were
used in the building of these models, though the machine-fit specifications typically returned models
that were far too overfit to the dependent variable.
Using these models, I generated monthly predicted probabilities of Republican and Democratic
control of the House and Senate, from 1940 to 2016. Figure A1 presents these extrapolations, as well
as the IEM market prices where available. In the figure, the light colors represent extrapolations of the
market prices, while the darker colors represent the actual IEM prices. Figure A2 presents the same
information, but it breaks down the projections by presidency rather than Congress. In both graphs,
the predicted probabilities comport nicely with contemporaneous reports about the upcoming election,
and they serve as the basis for assigning Congresses to the various Scenarios delineated in the theory
section.
As noted earlier, however, these majority control probabilities do not provide all of the predictions
necessary to measure which electoral scenario applies at a given point in time. In particular, they do
not cleanly translate into predictions regarding whether either party will capture the filibuster pivot. To
measure this probability, I build a third model, this time regressing actual seat share in election following
8

Table 5: Policy Deceleration (Scenario 1)
Dependent variable:
reasonable_prop
(1) (2) (3) (4)
SQ ∈Deceleration Region −0.839∗∗∗ −0.876∗∗ −1.990∗∗∗ −1.518∗∗∗
(0.289) (0.402) (0.364) (0.422)
Majority Status 2.148∗∗∗ 2.418 −2.029 −4.630∗∗∗
(0.741) (1.492) (1.250) (1.544)
SQ Location −0.078 0.008 0.002 −0.278
(0.071) (0.105) (0.173) (0.265)
|SQ Location| 1.907∗∗∗ 2.082∗∗∗ 2.469∗∗∗ 2.542∗∗∗
(0.166) (0.264) (0.243) (0.346)
Female −0.356 −1.197∗∗ −0.907∗∗ −0.736
(0.294) (0.519) (0.450) (0.688)
Democrat 0.766∗∗ 1.700∗∗∗ 2.159∗∗∗ 3.448∗∗∗
(0.324) (0.549) (0.653) (1.135)
Ideological Extremity −1.178∗∗∗ −1.054∗∗ −1.273∗∗∗ −1.689∗∗∗
(0.252) (0.483) (0.415) (0.551)
Committee Chair 0.133 −0.315 0.067 −0.363
(0.265) (0.370) (0.360) (0.480)
Ranking Member −0.263 −3.406∗∗∗ −0.428 3.802
(0.530) (1.214) (1.204) (8.827)
Subcommittee Chair −0.538∗∗ −0.590 0.244 −0.332
(0.246) (0.369) (0.326) (0.450)
Sub-Comm. Ranking Member 0.479 −0.312 −0.873 −2.063
(0.582) (1.277) (1.221) (2.054)
Seniority 0.00003 0.00004 0.00000 0.00004
(0.00003) (0.0001) (0.0001) (0.00006)
Constant −5.214∗∗∗ −6.094∗∗∗ −1.939 0.255
(0.946) (1.689) (1.570) (1.466)
Observations 753 445 507 320
Log Likelihood −283.181 −139.762 −152.904 −87.402
Akaike Inf. Crit. 592.362 305.525 331.808 200.804
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
9

Table 6: Policy Acceleration (Scenario 2)
Dependent variable:
Introduction of Viable Proposal
(1) (2) (3) (4)
SQ ∈Acceleration Region 0.737∗∗ 0.630 1.121∗∗ 0.695
(0.357) (0.488) (0.437) (0.548)
Majority Status 2.055∗∗∗ 2.438∗−1.475 −4.716∗∗∗
(0.747) (1.478) (1.186) (1.739)
SQ Location 0.024 0.106 0.154 −0.282
(0.064) (0.094) (0.143) (0.275)
|SQ Location| 1.907∗∗∗ 2.077∗∗∗ 2.284∗∗∗ 2.340∗∗∗
(0.165) (0.257) (0.229) (0.316)
Female −0.367 −1.195∗∗ −0.820∗∗ −0.723
(0.293) (0.524) (0.410) (0.640)
Democrat 0.917∗∗∗ 1.865∗∗∗ 2.274∗∗∗ 4.038∗∗∗
(0.332) (0.526) (0.580) (1.186)
Ideological Extremity −1.150∗∗∗ −1.089∗∗ −1.098∗∗∗ −1.560∗∗∗
(0.248) (0.465) (0.351) (0.529)
Committee Chair 0.148 −0.306 0.248 −0.301
(0.273) (0.376) (0.369) (0.469)
Ranking Member −0.215 −3.370∗∗∗ −0.198 5.515
(0.543) (1.193) (1.009) (882.746)
Sub-Comm. Chair −0.460∗−0.610∗0.187 −0.451
(0.242) (0.359) (0.314) (0.428)
Sub-Comm. Ranking Member 0.477 −0.222 −0.786 −1.808
(0.594) (1.279) (1.126) (2.228)
Seniority 0.00002 0.00003 −0.00002 0.00003
(0.00003) (0.0001) (0.00005) (0.0001)
Constant −5.674∗∗∗ −6.520∗∗∗ −3.427∗∗ −0.436
(0.966) (1.725) (1.433) (1.560)
Observations 753 445 507 320
Log Likelihood −285.914 −141.998 −169.325 −93.646
Akaike Inf. Crit. 597.827 309.996 364.650 213.292
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
10

Table 7: Policy Acceleration (Scenario 3)
Dependent variable:
reasonable_prop
(1) (2)
SQ ∈Acceleration Region 2.813∗∗∗ 2.467∗∗∗
(0.436) (0.584)
Majority Status 1.428∗1.340
(0.759) (1.289)
SQ Location −0.022 0.020
(0.080) (0.106)
|SQ Location| 2.237∗∗∗ 2.315∗∗∗
(0.182) (0.267)
Female −0.373 −1.299∗∗
(0.335) (0.589)
Democrat −0.276 0.543
(0.426) (0.671)
Ideolgoical Extremity −1.214∗∗∗ −0.992∗
(0.291) (0.516)
Committee Chair 0.408 −0.195
(0.267) (0.389)
Ranking Member −0.120 −3.721∗∗∗
(0.618) (1.302)
Sub-Comm. Chair −0.149 −0.545
(0.255) (0.406)
Sub-Comm. Ranking Member 0.604 −0.488
(0.628) (1.252)
Seniority 0.00000 0.00002
(0.00003) (0.0001)
Constant −5.718∗∗∗ −6.075∗∗∗
(0.997) (1.533)
Observations 753 445
Log Likelihood −248.755 −128.856
Akaike Inf. Crit. 523.509 283.713
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
11

Table 8: Predictive Models for Iowa Electronic Market Prices
Dependent variable:
Pr(Rep. House) Pr(Rep. Senate)
(1) (2)
Number of Current Republican Members 0.027 0.330∗∗∗
(0.018) (0.093)
Republican President and Majority 0.584
(2.508)
Senate Seats Defended −0.212∗∗∗
(0.071)
Presidential Approval 0.051 −0.007
(0.053) (0.041)
Democratic President 7.107 5.269
(5.130) (3.925)
Presidential Approval ∗Democratic President −0.091 −0.066
(0.087) (0.070)
Presidential Election Year −1.162 0.588
(1.100) (0.921)
Democratic President ∗Presidential Election Year −2.160
(1.414)
Congressional Time Trend 0.004
(0.054)
Generic Vote Poll Differential 2.132 0.066
(1.942) (0.062)
Congressional Time Trend ∗Generic Vote Poll Differential −0.020
(0.018)
Pepublican President and Majority ∗Presidential Election Year 2.684∗
(1.586)
Constant −10.217 −12.868∗∗
(7.022) (5.566)
Observations 92 92
Log Likelihood −26.069 −35.440
Akaike Inf. Crit. 74.138 88.880
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
12

Figure 1: Predicted Probability of Attaining Filibuster-Proof Majoirty in Senate
a
aNote: Democratic probabilities depicted in blue, Republican in red.
a given Congress on a variety of covariates similar to those presented in Table A8. In addition to these
covariates, I include the predicted majority control probabilities generated in Table A8 in the model.
Crucially, to estimate the model, I employ a Bayesian estimation of a Poisson count model, as doing so
allows me to calculate a distribution of predicted Senate seat counts for each observed combination of
covariate values. By calculating the percentage of this distribution for which either party is predicted
to attain 60 or more Senate seats, I can produce a probability that the filibuster pivot will be captured
by either Republicans or Democrats. is provides the final piece of electoral information needed to
assess which electoral scenario members face at a given point in time.
Table 9: Predictive Model for Control of Filibuster Pivot
Dependent variable: Republicans in Senate
2.5% 25% 50% 75% 97.5%
(Intercept) 2.8382887 2.9408208 2.9810628 3.026467 3.123543
IEM Projection (GOP) -0.1755138 -0.1170642 -0.0880277 -0.060353 -0.006238
GOP in Senate 0.0189805 0.0216515 0.0229784 0.024354 0.027087
Seats Defended by GOP -0.0215998 -0.0188637 -0.0172997 -0.015993 -0.013463
Gen. Vote Share Poll Differential 0.0025487 0.0034628 0.0039521 0.004442 0.005259
Presidential Approval -0.0002599 0.0005496 0.0009262 0.001387 0.002132
Democratic President 0.3032922 0.3691468 0.4024813 0.434416 0.494369
Presidential Election Year 0.0393485 0.0569597 0.0659358 0.075781 0.096964
Pres. Approval ∗Dem. President -0.0056338 -0.0047145 -0.0042138 -0.003619 -0.002686
Dem. President*Pres. Election Year -0.2463468 -0.2148064 -0.1975603 -0.183207 -0.158368
Observations = 38
Sample size per chain = 10000
inning interval = 1; Number of Chains = 1
13

E: Justification for Advancement Asymmetry
As noted in the main text, the agenda-setting game to which members respond features an asymmetry
in whether or not the game advances to a second round. In particular, the game only advances to
a second round if the SQ persists, forcing AS to make a decision between what she can achieve in
the present Congress, versus what she could achieve in future Congresses. Here, I provide additional
justification for why this design choice is appropriate.
In their recent paper on policymaking, for example, ?underscore the following commentary offered
by environmental advocates from eClean.org, regarding proposed cap-and-trade legislation that they
opposed: “Will [the public] see [the legislation] as a ‘win’—that the problem is solved? If so, what will
that mean for pushing for the needed steps later?” In other words, if compromise legislation prevails
today, such progress will preclude further reforms in the future. Policy advocates from other issue areas
echoed a similar sentiment in interviews for crosson2016working study on coalition lobbying, stating
that, “Passing legislation as close as possible to our ideal policy is important, because if we go back
to Congress next year and ask for the rest of what we want, they will deny us and tell us they have
already ‘done’ [issue redacted].”1In other words, Congress has already addressed the problem brought
forth by the public and interest community, and they must allocate scarce agenda space to some other
issue yet to be addressed. Consequently, policy entrepreneurs and interest group leaders understand the
importance of not squandering their window of opportunity by ceding too much to the opposition.
is observation also receives some support within the empirical literature. According to ?, a
large majority of major laws were not amended within four years of passage. In fact, only about
25 percent were amended within 1 to 3 years. Insofar as amendments serve as a good measure of
attempts to alter a targeted SQ more than once, these data would seem to confirm that repeated policy
movements within on the same, specific SQ issue area are rare. Taken together, then, both qualitative
and quantitative evidence provide a strong justification for the aforementioned asymmetry in game
continuation. Moreover, they provide some anecdotal evidence for the idea that political elites do in
fact consider future policy change possibilities in their present policy-change calculus.
It is worth noting that a possible addition or alternative to this design might be to tie electoral
fortunes to the majority party’s policy success in the present round. In other words, one may consider
1e specific issue area is here redacted due to IRB agreements to preserve the anonymity of interviewees.
14

endogenizing electoral outcomes to policy decisions in the present round. In this model, I do not
endogenize elections however, for a few key reasons. First, the conditions under which policy change
or stasis harm a majority party are unclear. If little policy change occurs, the majority party may be
punished for a refusal to compromise. If, however, the majority party does make major changes, they
may be punished for unpopular policies or poor outcomes (such as Democrats in the 2010 election).
Second, many gains and losses to a majority party’s seat holdings are cyclical and predictable. For
example, the president’s party typically loses seats in midterm elections. It is the effect of this sort
of predictable change, upon which majority party leaders can reliably condition their actions, that is
the focus of this study. Future work, however, may well examine how endogenous elections further
complicate majority agenda-setting and aggregate policy change.
15

F Regression Results with Member-Level Fixed Effects
Table 10: Policy Deceleration, Leveraging Only Within-Member Variance
Dependent variable:
Introduction of Viable Proposal
(1) (2) (3) (4)
SQ ∈Deceleration Region −0.914∗∗ −0.807 −4.233∗∗∗ −3.278∗∗∗
(0.450) (0.698) (0.950) (1.261)
SQ Location −0.129 −0.191 0.471 −0.250
(0.113) (0.191) (0.422) (0.653)
|SQ Location| 2.220∗∗∗ 2.281∗∗∗ 4.340∗∗∗ 3.967∗∗∗
(0.243) (0.365) (0.667) (0.909)
Constant −5.905∗∗∗ −6.508∗∗∗ −9.314∗∗∗ −10.323∗∗∗
(1.119) (1.543) (1.892) (3.052)
Member-Level FE Y Y Y Y
Observations 753 445 507 320
Log Likelihood −155.514 −68.375 −44.622 −22.890
Akaike Inf. Crit. 925.028 568.751 613.244 421.780
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
16

Table 11: Policy Acceleration (H2), Leveraging Only Within-Member Variance
Dependent variable:
Introduction of Viable Proposal
(1) (2) (3) (4)
SQ ∈Acceleration Region 1.677∗∗∗ 1.286 1.932∗∗ 0.952
(0.633) (1.069) (0.961) (1.511)
SQ Location 0.015 −0.111 0.324 −0.345
(0.109) (0.182) (0.378) (0.624)
|SQ Location| 2.260∗∗∗ 2.339∗∗∗ 3.503∗∗∗ 3.433∗∗∗
(0.244) (0.377) (0.515) (0.788)
Constant −7.351∗∗∗ −7.743∗∗∗ −9.695∗∗∗ −10.318∗∗∗
(1.286) (1.956) (1.956) (3.149)
Member-Level FE Y Y Y Y
Observations 753 445 507 320
Log Likelihood −154.136 −68.350 −58.263 −27.290
Akaike Inf. Crit. 922.272 568.699 640.525 430.579
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
17

Table 12: Policy Acceleration (H3), Leveraging Only Within-Member Variance
Dependent variable:
Introduction of Viable Proposal
(1) (2)
SQ ∈Acceleration Region 5.158∗∗∗ 5.171∗∗∗
(0.732) (1.061)
SQ Location 0.023 −0.050
(0.122) (0.212)
|SQ Location| 3.114∗∗∗ 3.019∗∗∗
(0.341) (0.496)
Constant −7.567∗∗∗ −7.818∗∗∗
(1.278) (1.745)
Member-Level FE Y Y
Observations 753 445
Log Likelihood −118.561 −50.175
Akaike Inf. Crit. 851.121 532.350
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
18