Stalemate in the States: Negative Agenda Control, Veto Players, and Legislative Gridlock in the American States

Full text

Stalemate in the States
Agenda Control Rules and Policy Output in
American Legislatures
Jesse M. Crosson∗
University of Michigan
March 1, 2018
Abstract
This paper examines how the power of majority party leaders to set the legislative
voting calendar influences policy change in American state legislatures. By generating
an opportunity for party leaders to exercise gatekeeping or negative agenda control,
such rules introduce an additional partisan veto player into a system of governance.
This addition typically increases the size of the core or gridlock interval, which drives
policy change downward. Using both traditional data on bill passage counts and new
data on Affordable Care Act compliance, I find strong support for these claims. More
specifically, when I calculate core sizes that are sensitive to agenda rules, I find that
core size is negatively correlated with policy change, as expected. Moreover, even when
I match states on their overall preference dispersion or polarization, the ability of party
leaders to exercise negative agenda control is strongly negatively associated with policy
change.
∗Ph.D. Candidate and Gerald R. Ford Fellow, Department of Political Science. Special thanks to George
Tsebelis, Rocio Titiunik, Chuck Shipan, Ken Kollman, Geoff Lorenz, James Strickland, and Eric Arias for
helpful feedback.
1

Over the past three decades, scholars of American political institutions have invested
much time and effort into exploring the causes and consequences of legislative gridlock.
Within the study of gridlock, however, few topics have generated the level of disagreement
as the role that political parties do or do not play in the policy change process. For some,
parties simply represent ideological coalitions, themselves contributing little to policy change
dynamics (Krehbiel 1998). For others, however, political parties are central to policy change,
as they exert a great deal of control over the legislative agenda (Cox and McCubbins 1993,
2005). Yet in spite of the fact that competing theories of political parties and policy change
generate specific, testable empirical implications, studies to date have often struggled to
delineate how much (if at all) political parties matter for policy change.
At least part of this struggle derives previous studies’ focus on policy change dynamics
in Congress alone. To be clear, insofar as the goal of these studies is to test whether
partisan agenda control occurs in Congress, focusing solely on the U.S. Congress makes
sense. However, as a means of testing the broader implications of partisan agenda control
for aggregate policy change, Congress has clear limitations as a setting for such examinations.
Among these limitations is the fact that most proponents of party-centric theories of Congress
argue that agenda-control developed as far back as the 1880s (Gailmard and Jenkins 2007)—
preceding the period over which empirical analysis is often conducted.
In this paper, I provide one of the first broad empirical documentations of the powerful
implications partisan agenda control has for aggregate policy change. To do so, I turn
to the institutional richness found in the American states and trace the influence of the
presence (and absence) of agenda control institutions through the policymaking process. In
doing so, I demonstrate that institutional features enabling negative or gatekeeping agenda
1

control significantly slow policy change, even beyond what preference polarization alone
might predict. More specifically, I find that 1) by increasing the size of the “core” or gridlock
interval,1the presence of partisan agenda control drives gridlock upward, and 2) even when
conditioning on distance between traditional institutional pivots, the presence of partisan
agenda control institutions negatively predicts policy change. Taken together, these findings
build upon Anzia and Jackman’s (2013) work on agenda control and roll rates and develop
support for the idea that negative agenda control introduces a new, partisan veto player into
a system of governance, thereby decreasing policy change. These findings also improve upon
earlier work on agenda-control in the states by Cox, Kousser, and McCubbins (2010), by
extending the analysis of agenda control past roll rates and individual policy shifts in two
states to aggregate-level policy change in across many state legislatures, from 1995-2014.
1 Legislative Gridlock: Parties and Preferences
The importance of political parties to policy change and legislative gridlock has long
remained a key topic of debate among legislative scholars. Indeed, beginning with Mayhew’s
(1991) extended exchange with Binder (1999, 2003) and others regarding the importance
of divided government, a great deal of scholarship has disputed whether and how political
parties contribute to legislative gridlock. But while much of the early discussion regarding
parties and policy change focused on divided government, Cox and McCubbins (1993, 2005)
and Krehbiel (1998) extend this discussion to the role that parties play within the legisla-
ture itself. According to Krehbiel’s account of the policymaking process, parties themselves
1The term “core” comes from Tsebelis (2002) and is more commonly used in studies of comparative
politics (though Tsebelis himself drew the term from Hammond and Miller’s 1987 analysis of the U.S.
Constitution). I use the term interchangeably with “gridlock interval” (Krehbiel 1998) in this analysis, as
the terms are conceptually identical, even though Tsebelis’s core can include partisan veto players (while
Krehbiel introduces no such players).
2

do little work to explain policy change dynamics, instead serving as ideological preference
aggregators. Consequently, it is the policy preferences of pivotal institutional actors—and
not party actors per se—that ought to influence policy change. In order to make predic-
tions about policy change, then, one should first delineate which actors are pivotal, and
then measure preference distances between them. The larger the distance between these ac-
tors, the larger the set of immoveable status quo policies—and the less policy change should
occur—regardless of an actor’s political party.
Cox and McCubbins (2005) provide a sharp response to this claim: far from mere pref-
erence aggregators, parties in Congress are instruments of reelection that create a well-
coordinated party brand. In order to maintain a healthy party brand, majority party leaders
are enfranchised with gate-keeping or “negative” agenda control, allowing them to prevent
bills that fracture the party from ever reaching the floor. Naturally, this conception of par-
ties and agenda control has strong ramifications for the amount of policy change a political
system ought to experience. Indeed, as I argue here (and as others, such as Chiou and
Rothenberg 2003 and Woon and Cook 2015 have explored), negative agenda control effec-
tively adds pivots to Krehbiel’s party-less model. That is, because of the majority’s desire
to allow votes only on legislation with a majority of their party’s support, chambers that
enfranchise the majority with negative agenda control exhibit a partisan pivot or veto player
located at the majority median—in addition to the usual pivot found at the chamber median.
Adding this pivot grows the theoretical gridlock interval, leading Cox/McCubbins’ model to
predict more gridlock than does Krehbiel’s: some policies that might pass under Krehbiel’s
model would not even receive consideration for a vote under Cox/McCubbins’ model.
Several studies have attempted to determine empirically whether and how majority par-
3

ties exercise negative agenda control, most often focusing on majority roll rates (e.g., Anzia
and Jackman 2013, Lawrence, Maltzman, and Smith 2006). Others have attempted to ad-
judicate between the Krehbiel and Cox/McCubbins frameworks by following each theory
through the policymaking process and examining which model better predicts observed lev-
els of policy change, using a variety of innovative methodologies (e.g., Chiou and Rothenberg
2003, Richman 2011, Peress 2013, Woon and Cook 2015). Most find support for some kind
of gatekeeping role for majority leaders in Congress.
Each of these studies make important contributions to the study of how parties operate
in Congress. However, as examinations of agenda-control institutions, their implications,
and their general effects, they are limited by their sole focus on Congress.2First, as an
institutional setting for studying agenda control, Congress presents a variety of empirical
challenges. For instance, congressional data are sometimes poorly suited for making fine
empirical distinctions between models of policymaking: insofar as analyses focus on the last
100 years of congressional history, they are confined to an era in which agenda control in-
stitutions are not thought to have varied.3Relatedly, within a single institution such as
Congress, predictions from differing theoretical models sometimes turn out to be observa-
tionally equivalent, as Woon and Cook (2015, p. 1) underscore. Additionally, because of the
difficulty of collecting historical data, congressional studies are handicapped by small sam-
ple sizes, rendering fine distinctions between theoretical predictions even more difficult (e.g.,
Peress 2013). But beyond the empirical limitations implied by a singular focus on Congress,
federal-level work also typically focuses on determining which models fit policy change data
2Anzia and Jackman (2013) and Cox, Kousser, and McCubbins (2010) provide notable exceptions.
3Anzia and Jackman (2013) make a similar point regarding party power in Congress versus the states.
4

best, rather than tracing the overall effect of partisan agenda control on policy change. In
this way, previous studies teach us a great deal about the inner-workings of Congress, but
less about the broad ramifications of partisan agenda control for a system of governance.
I therefore focus my analysis on the American states. Unlike Congress, state legislatures
vary considerably in the presence of partisan agenda control institutions. This variation
(and much larger sample size) generates an excellent environment to examine how partisan
agenda control (and the exercise of negative agenda control specifically) is tied to policy
output. Within this context, I can measure directly whether or not individual chambers in
each state possess partisan agenda control institutions, and then trace the overall infuence
of these institutions on policy change. In this way, I build upon work by Cox, Kousser,
and McCubbins (2010), who determine that the introduction of agenda-control institutions
in state legislatures (namely, Colorado and select bills in California) drastically influences
majority roll rates and the direction of policy shifts in those states.
Beyond the empirical contributions of this approach to the study of negative agenda
control and its ramifications, my approach also contributes new knowledge to present work
on policy change in state legislatures. To date, research on policy change in the states
has examined a number of factors that are distinctive from national-level studies of policy
change. Gray and Lowery (1995) and Bowling and Ferguson (2001), for example, examine
the influence of interest-group density and diversity on legislative gridlock. Gray and Lowery
find that the number of interest groups positively influences legislative productivity, while
Bowling and Ferguson find that interest diversity stymies policy change. Additionally, Rogers
(2005) and Hicks and Smith (2009) examine how factors such as term limits and direct
democracy might influence policy change at the state level. In spite of these advances,
5

few if any studies have examined how pivotal actors’ preferences influence policy change in
the states—much less how partisan institutions might do so. By examining these factors,
this study contributes to our current understanding of policy change dynamics in state
legislatures, in addition to its contribution to current work on partisan agenda control.
2 Theoretical Expectations: How Does Negative Agenda Control
Work to Slow Policy Change?
In general, negative agenda control is defined as the ability of an actor to keep an item
from receiving a final decision, most commonly a vote (e.g., Cox and McCubbins 1993,
2005, Gailmard and Jenkins 2007). In the legislative context, majority parties are thought
to exercise negative agenda control by disallowing votes on legislation that fractures the
party caucus. Here, I focus on negative agenda control wielded by a chamber’s floor leaders
via their ability to set the voting agenda as bills emerge from committee. I focus on this
institutional feature because it captures well the concept of partisan gatekeeping, and because
of its apparent effectiveness at slowing roll rates according to previous research (Anzia and
Jackman 2013, Jackman 2013, Cox et al. 2010). Indeed, when party leaders can prioritize
and deprioritize legislation on the floor voting agenda, they may effectively avoid votes that
roll the majority party. Conversely, when the voting calendar is set by an “automatic” rule,
roll rates are found to be considerably higher. Such automatic rules include “first-come-first-
served” arrangements, as well as the use of alphabetical order by sponsor name.
I argue that the presence of such partisan agenda control introduces a new, partisan veto
player into a system of governance, located at the median of the majority party. That is, if
the median of the majority party (and therefore a majority of the majority party) disapprove
of a piece of legislation, it will not receive a vote in a chamber with partisan gatekeeping.
6

This addition should increase policy stability, for as Tsebelis (2002) demonstrates, the more
veto players in a lawmaking system, and the more preference distance between them, the
larger the “core”, or set of status quo policies that cannot be defeated by any policy proposal
in a political system.4The larger the core or gridlock interval, the more policy stability a
system should encounter on average.5
However, additional veto players do not always increase policy stability: if a veto player is
“absorbed” by another veto player, that veto player will not increase policy stability. A veto
player is absorbed if she, by virtue of her preference set, would not choose to independently
veto a change to the status quo. Put differently, in the cases for which an absorbed veto player
might choose to veto legislation, there will always exist another veto player who also would
choose to veto that piece of legislation. In the context of American legislatures, a moderate
Republican executive may be absorbed in a political system that features one Democrat-
controlled legislative chamber and one radical Republican-controlled chamber. While such
an executive may object to, say, a liberal piece of legislation originating in the Democratic
chamber, the radical Republican chamber will also oppose such legislation. In this scenario,
the Republican executive, as an absorbed veto player, has no incentive to unilaterally veto
any proposal and does not add to the political system’s policy stability.
4Tsebelis defines the core similarly to how Krehbiel (1998), Chiou and Rothenberg (2003), and many
others define the gridlock interval: the core is the set of status quo policies that cannot be defeated by
any policy proposal. Clinton (2012, 88) similarly defines the gridlock interval as the “regions in the policy
space where status quos either cannot or will not be changed according to the model.”’ The primary
difference between Tsebelis’s model and the model that Krehbiel details is that, while Krehbiels model is
one-dimensional, Tsebeliss model generalizes to n policy dimensions.
5Note that these predictions regarding gridlock interval size and policy change derive from static models
of policy change. In each of these models, Nature selects a status quo policy for consideration by the relevant
veto players. Consequently, while the size of the gridlock interval generally should correlate with negatively
with policy change, the distribution or “supply” of status quos could in practice change over time and
thereby influence the observed amount of policy change. I thank an anonymous reviewer for underscoring
this challenge, which I address below.
7

Most American legislatures are thought to have three main institutional veto players: a
lower chamber median, an upper chamber median, and an executive.6Given these similarities
across state governments, it is theoretically true by construction that the negative agenda
control core therefore “covers” the institutions-only core that incorporates chamber medians
alone. However, because I leverage cross-state variation in the presence of negative agenda
control in this state, core size may not necessarily correlate strongly the presence of negative
agenda control, due to absorption. That is, if states without negative agenda control were
to exhibit more overall preference polarization than states with negative agenda control,
the correlation between agenda-control institutions and core size would be weak or non-
existent. Such a phenomenon would restrict one’s ability to use cross-state data to examine
the influence of negative agenda control on policy change.
Thus, it is important to test whether such a correlation exists. This claim is summarized
in H1:
H1: American state legislatures with majority parties that exercise negative
agenda control should, ceteris paribus, have larger gridlock intervals than do
those without such parties (and, consequently, fewer veto players).
Given that H1 is true by construction within any given state, I relegate tests of H1 to
Online Appendix A. Indeed, even though cross-state differences in preference polarization
may obfuscate the relationship between agenda-control institutions and core size, H1 serves
more as a data “check” than a traditional hypothesis test.7Within Online Appendix A, I
6Nebraska’s unicameral legislature serves as a clear exception.
7Another way to think of H1 is through the framework of experimental research. If negative-agenda-
control institutions serve as the “treatment” in the study, testing H1 is akin to an experimental researcher
ensuring that treated units did not fall victim to treatment failure. In the legislative context, H1 ensures
8

find support for H1: the presence of negative-agenda-control institutions is indeed positively
associated with core size.
Given this finding regarding core size, the data therefore are suitable for testing the first
of two hypotheses concerning negative agenda control, the core, and legislative gridlock:
H2: The larger the agenda-control-adjusted core, the more gridlock (less policy
change) a system will encounter.
Finally, in addition to demonstrating that the agenda-control-adjusted core best predicts
policy change, my analysis aims to show that the presence of partisan agenda control insti-
tutions consistently predicts differences in policy output between states that are otherwise
similar in their levels of preference polarization. To do so, I will show that negative agenda
control does indeed matter for gridlock, even when accounting for the size of a Krehbiel-like
gridlock interval:
H3: Even conditional on distance between institutional veto players, negative
agenda control should lead to higher levels of gridlock in states with legislative
majorities that possess it.
3 Agenda Control, Adjusted Core Size, and Policy Change
In order for agenda-control institutions to influence policy change, I argue that they do so
by increasing the size of the core or gridlock interval (H2). I test this central expectation in
two ways: first by using the most widespread measure of legislative productivity in studies of
state legislatures (bill passage counts), and then using a new measure of policy change based
that the “treatment” is positively associated with the theorized causal mechanism, core size, thought to
influence policy change.
9

on Affordable Care Act (ACA) implementation. First, however, I detail how I measure core
size in each state-year, accounting for whether or not a state possesses legislative institutions
enabling partisan agenda control.
Measuring the Agenda-Control-Adjusted Legislative Core
In order to measure the agenda-control-adjusted core, I first determine whether or not
a chamber’s majority party may wield negative agenda control, using information found in
Anzia and Jackman’s (2013) replication data. More specifically, I measure whether majority
leaders enjoy control of the legislative floor-vote calendar. Under this rule, majority party
leaders decide which bills, among all of the bills that could come up for a vote, will actually
receive a floor vote—and when. Using the presence (or absence) of this rule, I code each leg-
islative chamber in a binary fashion, with a chamber taking on a value of 1 (majority party
possesses calendar control) or 0 (majority party lacks such control). When pooled across en-
tire legislatures, this variable may therefore take on three possible values: 2 (negative agenda
control in both houses), 1 (control in just one chamber), or 0 (no negative agenda control).8
Given that all American states (besides Nebraska) have the same number of institutional
veto players, differences in this negative agenda control variable will also represent overall
differences in a state’s total number of veto players.
Core or gridlock interval size, however, is a function of more than just the number of
veto players in a system: preference distance between veto players also influences the size of
the core (Tsebelis 2002). Thus, to measure the size of the gridlock interval, one first needs a
measure of veto player preferences. In this study, I utilize Shor and McCarty’s (2011) NPAT
8As noted earlier, when a majority party lacks calendar powers, the agenda is most commonly set by
automatic rule—very often a “first come, first served” arrangement.
10

scores to measure these preferences. As Clinton (2012) states, using roll-call-based measures
of preferences to test partisan theories carries with it potential limitations. In spite of these
limitations, however, Shor and McCarty’s scores provide the best means for measuring state-
level gridlock intervals, which have yet to be examined at the state level. In using these scores,
I assume unidimensionality of issue space in the state legislatures. Determining gridlock
interval size in unidimensional policy space is quite straightforward: once one determines
the number and position of veto players in a system of governance, unidimensional core size
is simply the maximum distance between any two veto players. Thus, calculating core size in
each state proceeds by first determining how many chambers (if any) have majority parties
exercising negative agenda control, then measuring the locations between each relevant veto
player, and finally selecting the maximum distance between veto players in the state-year
in question.9In other words, instead of identifying which of the intervals will serve as the
core, the number of veto players simply identifies which veto players (and therefore which
set of preference distances) must be maximized over in order to calculate the core. Here, one
should note that, were ideal points for governors available, distances between the governor
and relevant legislative veto players would be included among this set of distances to be
maximized over. However, given the that such scores are unavailable, I incorporate the
9To be clear, this strategy does not take into account the presence of filibuster and override pivots. There
are two primary reasons for this exclusion. First, with regard to filibusters, the presence and location of a
filibuster pivot is not as clearly defined in the states as it is in the U.S. Senate. While this problem is further
detailed in Online Appendix E, there is limited consensus on which states “have” a filibuster pivot: while
some states appear to have rules akin to cloture, the vast majority of states have either strict, automatic
limits on speech, or no rules pertaining to speech whatsoever. Moreover, these rules speak little to the actual
practice of filibustering and invocation of cloture in state legislatures. These challenges notwithstanding,
I present models in Online Appendix E that attempt to control for the presence of a filibuster pivot in a
legislature. Substantive results remain unaffected. With regard to veto override pivots, I argue that such
actors are often absorbed and thereby do not decrease policy change. Moreover, without a measure of the
governor’s preferences, it is difficult to know a priori whether to include the override pivot or governor in
core calculations, as both cannot simultaneously be pivotal.
11

preferences of the governor via a Divided Government measure, detailed below. Each of
these measurements are calculated biannually for each state, from 1995 to 2014.
[FIGURE 1 HERE]
Figure 1 provides greater detail on this process. In theory, chambers exhibiting partisan,
negative agenda control carry with them the potential for much larger cores than the average
chamber lacking such a feature, because agenda control increases the total number of veto
players. In Cases 3 and 4 of Figure 1, for example, majority parties in both chambers
exercise negative agenda control. Were legislatures in such states exposed to a divided
legislature, Cases 3 and 4 should produce a larger core than those depicted in the other cases
in Figure 2. The same logic applies for Cases 5 and 6 (just one chamber with negative agenda
control), which should create a larger core than Case 7 (no negative agenda control), all else
equal. Legislatures with larger numbers of veto players do not always possess larger cores,
however, due to the logic of absorption highlighted above. Cases 1 and 2 (two chambers
with negative agenda control, controlled by the same party) provide an example. Here,
the distance between the party medians is zero—the majority parties in these chambers
are very similar. Instead, the relevant gridlock interval or core distance is a tie between
|Mhm −Chm|,|Msm −Csm |,|Mhm −Csm|,and |Msm −Chm |(the distances between the
lower house majority party and chamber medians, upper house majority party and chamber
medians, lower house majority party and upper house chamber medians, and upper house
majority party and lower house chamber medians, respectively).
In measuring the core in this fashion—the maximum distance between relevant veto
players—I build on the approach developed by Krehbiel (1998), Chiou and Rothenberg
12

(2003), and others.10 That is, I compare the size of the gridlock interval to levels of policy
change over time. It is important to note, however, that this approach carries with it a
key assumption regarding the distribution of status quo policies. More specifically, I adopt
the assumption that status quo policies are drawn from a uniform distribution, just as
Krehbiel (1998) and Chiou and Rothenberg (2003) do. Thus, all results presented in this
paper should be interpreted with this assumption in mind, as it renders the results most
directly comparable to studies that make similar assumptions (e.g., Covington and Bargen
2004, Chiou and Rothenberg 2003, 2006, and 2009). However, as Tsebelis (2002), Clinton
(2012), and Krehbiel (1998) suggest, this assumption may not always hold true, particularly
if the previous legislature drastically changed the supply of alterable status quo policies.
Consequently, I present in the Online Appendix a series of robustness checks that control
for some of this potential variation. The results presented in this section are robust to each
check. Moreover, as detailed below, the results are also substantively similar when I use a
dependent variable that is more sensitive to the distribution of the status quo.
Test 1: Measuring Policy Change Using Bill Enactments
Using these core size measurements, I test H2 using two different outcome variables: one
that draws from previous work on state-level gridlock, and another that addresses common
concerns with the traditional approach. To date, models of legislative gridlock in the Amer-
ican states have frequently relied upon raw bill enactment counts as a means for measuring
gridlock (see, for example, Bowling and Ferguson 2001, Gray and Lowery 1995, Hicks and
Smith 2009, Rogers 2005). As Tsebelis, Binder, and Mayhew (among others) each detail,
10My approach is also similar to Tsebelis’s (2002), though there are some differences due to the wider
availability of data and measurements available today.
13

this measure possesses a number of important flaws. Nevertheless, due to the difficulty of
collecting data on “significant” legislation in all 50 state legislatures, studies have neverthe-
less focused on raw productivity numbers. Thus, I first test H2 with these measures, in order
to relate my findings to these studies of gridlock. For these tests, bill enactment counts are
taken from the Council on State Government’s Book of the States, for the years 1995-2014.
Under this measurement, the more bills passed, the greater the movement of the status quo.
Before examining the relationship between core size and bill passage, a final aspect of core
size merits attention. That is, as noted above, core size does not pertain just to legislative
veto players: because of the veto power of governors, the location of the governor matters
for the size of the core. Given that Shor and McCarty’s ideology scores do not extend
to governors, I introduce a divided government variable into my empirical models. The
variable takes on the value “1” if the governor is of a different party than both chambers of
the legislature and “0” otherwise.11 I expect that, ceteris paribus, divided government will
positively correlate with gridlock.
I estimate a variety of models to test the robustness of the relationships in the data.
First, I estimate a simple negative binomial model. Second, to better account for unmeasured
factors specific to each state and year, I estimate models using state- and year-specific effects.
Because such effects present challenges to estimating negative binomial models, I therefore
use logged bill passage counts as the dependent variable in these models and then estimate
the models using OLS. In these models, I use state random effects and year fixed effects,
11Divided government is calculated in this way because, while legislative elections occur at the district
level, governors are elected at the state level. As a result, governors are likely more moderate than legislators,
rendering them absorbed by any system that features one Democratic and one Republican legislative chamber
(per the example of absorption articulated above).
14

along with robust standard errors.12 Finally, I estimate linear models of logged enactments,
using panel-corrected standard errors alone. In each model, data span from 1995 to 2014,
and core size and enactment data are grouped biannually, to account for legislatures that
meet only once every other year. In addition to the core size and divided government
covariates noted above, I also include variables found in previous models of bill enactments:
state’s gross domestic product, interest group population, number of initiatives, number of
bill introductions, legislative professionalism, partisan dominance, state population, average
legislator ideologies in the upper and lower chambers, and total number of legislators.
With regard to State GDP, I expect that all else equal, states with larger economies
should exhibit a higher number of enactments. I anticipate this variable to function similarly
to another variable in the model, State Population. Moreover, I anticipate the larger bill
“supply” associated with Bill Introductions and Professionalism should drive enactment
counts upward. Conversely, Number of Interest Groups should be negatively associated with
policy change. As Gray and Lowery (1995) argue, a more crowded interest environment
can stymie policy change. Additionally, insofar as the number of initiatives indicates the
ease of the initiative process, I anticipate that Initiatives will be negatively associated with
enactments (see Gerber 1996 and Tsebelis 2002). Finally, I include Average Ideology of the
upper and lower chambers in order to account for the possibility that conservatives prefer
less policy change overall.
[TABLE 1 HERE]
Table 1 summarizes the results. As predicted, core size is negatively associated with
12A Hausmann test affirmed the choice to include state-level random effects in the model, instead of fixed
effects.
15

bill passage rates across all model specifications. In fact, holding other variables at their
means, a shift from the 10th percentile in Size of Core to the 90th percentile corresponds
with a predicted reduction in output of nearly 140 enactments (706 to 566, good for a ≈20%
reduction). Without state-level effects, the results are significant at the p < .01 level; with
state-level effects, the estimates remain significant (papproximately 0.05). Still, a reason-
able objection here may be that core size is simply proxying for divided legislatures: that
is, an institutions-only (i.e., chamber-medians-only) gridlock interval could explain policy
change just as well, and states with negative agenda control may just be correlated with the
presence of a divided legislature. In supplemental analyses, however, I find this not to be
the case. That is, when I substitute the agenda-control-adjusted core for other preference
distances (such as the distance between chamber medians), no other preference distance but
“Size of Core” reaches statistical significance. In other words, the agenda-control-adjusted
core is uniquely able, compared to the institutions-only core, to predict policy change in a
statistically significant fashion. This result holds for both this dependent variable, and the
significance-adjusted variable presented below.
Interestingly, Divided Government does not negatively predict enactments at a statis-
tically significant level in any of the models. Initiatives, State GDP, and State Population
also exhibited fairly inconsistent results, failing to reach statistical significance in most cases.
Moreover, Number of Interest Groups exhibits the predicted negative relationship in just one
of the four models. Bill Introductions, on the other hand, behaves as expected.
A few other results from the regression bear mentioning. First, Professionalism unsur-
prisingly falls out of significance when the panel structure of the data is accounted for (using
either panel-corrected standard errors, or more importantly, state-level effects). This is likely
16

due to the fact that the data present little within-state variation in professionalism during
this time period. Interestingly, though, in Model 1, professionalism is strongly negatively
correlated with enactments—in spite of the fact that professional states are often more popu-
lous and have larger economies than unprofessional ones. Additionally, Partisan Dominance
is strongly positively associated with enactment rates. Perhaps counterintuitively, the more
dominated a state is by a single party, the more productive they appear to be.
Taken together, these results are consistent with H2, that agenda-control-adjusted core
size does indeed decrease policy change.13
Test 2: ACA Implementation as a Measure of Policy Change
While enactment counts provide suggestive evidence of a negative relationship between
core size and policy change, such counts exhibit several shortcomings as a measure of policy
change. Indeed, many studies of policy change and gridlock have avoided measuring policy
change via enactment counts, due primarily to concerns about bill significance. Enactments
relay only limited information about status quo movements: one government could pass
10 incremental laws that do not move the status quo as much as another government’s
single piece of landmark legislation. Thus, Tsebelis, Mayhew, and others have chosen not to
measure policy stability via raw bill counts and instead focused on major enactments.
Given these challenges, I retest H2 taking policy significance into account. To do so, I
follow Tsebelis (2002) and focus on a single policy area: health policy. In particular, I focus
on implementation of the Affordable Care Act in the states, from 2011 to 2013. Focusing on
this specific policy area offers a number of benefits. First, it allows for better data collection
13It should be noted that, although California is an outlier in many of the enactments models, the results
presented here are robust to exclusion of California from the models. Results in the following empirical tests
also behave similarly, even when California is excluded from the analysis.
17

regarding bill significance. Second, it offers a partial means for dealing with the problem of
the status quo. Indeed, just because a government does not move the status quo does not
mean that it is unable to do so. Rather, relevant veto players may simply prefer the status
quo to feasible alternatives. By subjecting all 50 states to all implementation requirements
and incentives, the ACA shifts the health policy status quo in such a way that all 50 states
must respond in some way (even if only to deny funds or offer ACA alternatives).
In order to measure bill significance in this context, I utilize the National Council of States
Legislatures (NCSL)’s database on ACA-related bills in the state legislatures. NCSL has
developed 10 policy categories for ACA-related bills, with which they classify bills according
to relevant categories. For instance, NCSL would categorize a bill twice that addresses both
Medicaid and insurance exchanges: once in the Medicaid/CHIP category, and once in the
insurance exchange category. Classifying bills in this manner offers a means by which one
may create a measure of bill significance: because significant bills are “double-counted” (or
triple- or quadruple-counted) under this classification scheme, one may utilize the sum total
of relevant bills in each category as a means of measuring a state’s ability to move the status
quo on ACA compliance. This, then, is the approach I use to measure policy change and
gridlock. For each legislature, I count the total number of bills listed in each NCSL ACA
implementation category. Subsequently, I add together the bill counts from each category to
arrive at a single count for each state. This number represents the total “amount” of ACA
Implementation passed in each state, as the double-counting of more complex legislation
allows it to capture bill significance in a way that raw bill counts cannot.1415
14The figure in Online Appendix F depicts the search interface used to collect these data, along with an
example of how the dependent variable was calculated.
15These are cross-sectional data and are calculated by adding up all enactments over the three-year period
covered by the data (2011-2013).
18

Using this measure of policy change, I retest the relationship between core size and policy
change (H2). The data again provide support for the idea that core size drives policy change
downward.16 Beyond the covariates found in the raw bill enactment models, I include in
the ACA compliance models the partisanship of the state’s governor. I do this to account
for differences in ACA compliance due to distaste for or strong opposition to the ACA (a
controversial bill, passed entirely by Democrats). Table 2 summarizes the results. Here, I
estimate a negative binomial model, using clustered standard errors by state. Most variables
behave as expected. Here again, core size is again negatively associated with policy change
(p < .05). Indeed, even when adjusting for bill significance, a larger core size is associated
with less policy change. The same may be said about Divided Government, which is (mod-
erately) negatively associated with policy change (p < 0.1). In addition to these findings,
the model offers support for the idea that states with Democratic governors and states with
larger economies experience more policy change under this ACA measure. The former is
perhaps not surprising, as one particularly important aspect of ACA implementation, Med-
icaid expansion, vested a great deal of implementation power with governors. Interestingly,
though, the model’s measure of legislative liberalism, Mean Core Ideology, did not reach
statistical significance in these models. The same may be said about Partisan Dominance.
[TABLE 2 HERE]
These results again provide support for my theoretical expectations regarding core size
and gridlock. Larger core sizes appear to be associated with smaller changes to the status
quo. Figure 2, a marginal effects plot of core size and predicted policy change, captures the
estimated magnitude of this relationship. There, move from the 10th percentile to the 90th
16It should be noted that H1 was also retested on these data from 2011-2013 and again received support.
19

percentile in core size corresponds with a decrease of approximately 6-7 units of ACA status
quo movement—equivalent to a roughly 42 percent reduction.
[FIGURE 2 HERE]
Yet while these results are consistent with expectations, these models are not entirely
able to address the role that negative agenda control itself plays in the increase of gridlock.
Thus, the final analysis attempts to establish that the addition of a partisan veto player via
negative agenda control is responsible for higher levels of gridlock in state governments.
4 Does Negative Agenda Control Influence Policy Change Apart
from Institutional Polarization?
Section 3 appears to establish a connection between the agenda-control-adjusted leg-
islative core and the amount of policy change a political system experiences. However, as
Krehbiel (1993) argues, those findings may simply be a relic of overall preference polarization:
perhaps it is the case that states with agenda control institutions happen to be more polar-
ized or are more likely to have divided legislatures. In order to assess the impact of negative
agenda control on gridlock, then, one needs account for how polarized institutional pivots
are, and then assess whether or not negative agenda control displays a negative relationship
with policy change (H3). I undertake a matching analysis to address this challenge.17 In
addition to the benefits typically attributed to matching, such as an avoidance of structural
interpolation and a sensitivity to data (non)overlap, matching suits H3 particularly well.
Indeed, H3 claims that even when conditioning on institutional distances and other charac-
teristics that matter for gridlock, systems with negative agenda control still exhibit lower
levels of policy change (higher levels of gridlock) than do systems without negative agenda
17I present additional empirical tests in the Online Appendix.
20

control.
Here, I examine one primary type of treatment: I match and compare systems with
any amount of negative agenda control (Wi= 1) with systems lacking any kind of negative
agenda control (Wi= 0). In order to match treated and control units, I aim to condition
on an X vector that includes: 2008 presidential vote share, party of the governor, divided
government, size of state economy, and number of interest groups.18 Importantly, this Xalso
includes five different preference distances: |Mhm −Msm|,|Mhm −Csm|,|Msm −Chm |,|Mhm −
Chm|,|Msm −Csm |,and |Chm −Csm|(see Figure 1). These distances capture all the possible
inter- and intra-chamber combinations of institutional and partisan veto players in a system
with the maximum number of veto players. Of course, not all of these distances should matter
in all states. However, conditioning on these distances ensures that I am matching state
governments with maximally similar ideological spacing. If differences are found between
cases matched on these distances, even when some distances are (theoretically) irrelevant
in a given state, such a result would (and does) suggest that negative agenda control itself
impacts gridlock, beyond the impact of preference polarization alone.
Before discussing the results from this analysis, however, it is important to underscore
that this test is not meant to establish an effect of negative agenda control apart from
its impact on the size of the core or gridlock interval. Rather, it is meant to address the
possibility that the demonstrated correlation between policy change and my agenda-control-
adjusted gridlock interval is simply an artifact of preference polarization. As noted above,
proponents of partyless, preference-only theories of policy change contend that policy change
18For robustness, I also match on the three filibuster variables found in Online Appendix E, with results
strengthening upon inclusion of those variables.
21

can be explained on the basis of preference dispersion among institutional pivots alone.
Therefore, a reasonable objection to the H2 analysis might be that the results are driven
primarily by cross-state differences in overall preference dispersion in the legislature. In
response, this test (and an additional test in the Online Appendix) seeks to demonstrate
that negative agenda control (and the size it adds to the core) slows policy change, above
and beyond what preference polarization alone might predict. To this end, the matching
analysis conditions on two different kinds of preference dispersion: institutional polarization
(derived from divided legislatures) and partisan polarization (large distances between the
political parties).19 Should negative agenda control demonstrate a negative assocation with
policy change even after matching on these preference dispersion measures, the results would
suggest that negative-agenda-control legislatures are not simply more polarized.
Below, I estimate both the average treatment effect on the treated and average treatment
effect on the controls:
AT T =E[Yi(1) −Yi(0)|Wi= 1]; AT C =E[Yi(1) −Yi(0)|Wi= 0]
I argue that my analysis meets the requirements of strong ignorability and adhereance to the
Stable Unit Treatment Value Assumption (SUTVA), required of matching analysis. SUTVA
is violated if the assignment of negative agenda control in one state affects the outcome
(policy change) in another. The map in Figure 3 suggests that geographically similar states
have similar treatment assignments; however, it does not necessary suggest that a given
19As noted above, I match on both instituional an partisan polarization simultaneously. However, in
Online Appendix C, I present matching results for which I match on institutional polarization (distance
between chamber medians) and partisan polarization (distance between party medians) separately. In both
bases, the results remain substantial and significant, as in the analysis below.
22

state’s assignment affects that state’s outcomes. Instead, the presence of negative agenda
control in, say, Arizona, would have to influence policy change in, say, Nevada, in order for
SUTVA to be violated. This, of course, is entirely possible: if Arizona is unable to overcome
negative-agenda-control-induced gridlock and pass something like a time-sensitive tax policy,
Nevada may more vigorously pursue such policies, perhaps to attract regional businesses.
[FIGURE 3 HERE]
For a number of reasons, however, this concern is a minimal one for this study. First, every
state has to comply with the ACA, so the aforementioned race is unlikely to occur. Second,
because of absorption, negative agenda control may not affect gridlock so drastically that it
would become noticeable and influence a neighboring state’s actions. In the first place, many
states have possessed their negative agenda control rules for decades—long before current
legislative leaders ever took office. Thus, it would be difficult to attribute differences in
gridlock to the presence or absence of negative agenda control. Moreover, even if agenda-
setting powers are present in one state and not the other, negative agenda control need not
lead to gridlock. If it doesn’t, then it is unlikely that legislative production in the neighboring
state will respond in any meaningful way to the presence of a single procedural rule.
In addition to SUTVA, the assumption of unconfoundedness also merits attention. That
is, after conditioning on X, outcomes should be orthogonal to treatment assignment. While
it is impossible to test for the presence or absence of unobserved confounders, I attempt to
account for a wide variety of covariates commonly found in studies of legislative gridlock. By
conditioning on these common determinants of gridlock, confounding from these variables
should be accounted for. Finally, matching analysis rests crucially on the presence of overlap
in each of the dimensions of X. That is, the conditional distribution of controlled units ought
23

to share a common support with treated units on pertinent covariates. Figure 4 examines
the overlap assumption graphically. Observing the treatment versus control density plots for
each of the elements of X, the data in this study appear to possess healthy levels of overlap
overall. In fact, the covariate with the weakest balance, number of interest groups, generates
at-test that does not achieve significance at the p<.10 level. Table A7 in the Online
Appendix provides more detailed balance statistics, and Table 3 lists the actual matched
cases in the treatment and control groups. When combined with the unconfoundedness
assumption, this presence of overlap suggests that the data in this study meet the criteria
for strong ignorability necessary for matching analysis.
[FIGURE 4 HERE]
[TABLE 3 HERE]
In order to match treated and control units, I implement nearest neighbor matching, using
the GenMatch() function in R (Diamond and Sekhon 2013). With these matches, I estimate
the ATT and ATC. As Table 4 indicates, a difference-in-means t-test between treated and
control groups is statistically significant (p<.05) and substantial. The approximately 13-
unit difference points in the expected direction and represents a wide divergence in ACC
implementation (which ranges from a minimum of 0 to a maximum of 87 moments) among
treated and control units. This result is consistent with the claim that negative agenda
control does contribute to gridlock in the American state legislatures.20 21
20It should be noted that this result has been substantiated with tests conducted on the bill enactments
data. In the first such test, reported in Online Appendix C, I substitute the ACA measure of policy change
for enactments, and then rerun the matching analysis. The results are again negatively and highly significant,
with a difference in means of nearly 400 bills—a substantial difference.
21These results also hold if a modeling approach is instead taken. In Online Appendix C, I regress both
enactments and the ACA policy change measure on all of the aforementioned preference distances, along
with a variable indicating the number of chambers with party leaders controlling the calendar. In spite of the
inclusion of all possible preference distances, the agenda control variable remains negatively and statistically
24

[TABLE 4 HERE]
Online Appendix C also features an alternative test of H3 that considers directly the
mechanism by which negative agenda control ought to influence policy change (core size
increase). There, I decompose the core into an institutions-only component, including the
distance between legislative chambers as its own term in a regression model of enactments
and ACA implementation. Because the institutions-only core is by definition a subset of the
agenda-control-adjusted core, I then include a separate term in the model that represents
the “added distance” attributed to the core by the presence of negative agenda control
institutions. For both dependent variables, this added distance negatively and significantly
predicts policy change, as expected. Moreover, in all cases, model fit improves upon inclusion
of this added distance term. Thus, like the matching analysis, this test lends support to H3:
that negative agenda control contributes to policy change, above and beyond the level of
policy change predicted by institutional pivots alone.
5 Implications and Conclusions
Policy change and stability matter in a representative democracy. Indeed, while some re-
search shows that policy stability allows the economy to grow and prosper (e.g., Henisz 2000,
Acemoglu and Johnson 2005), many other studies also document the negative ramifications
of a legislature’s inability to address a polity’s problems (e.g., Alesina and Drazen 1991,
Mann and Ornstein 2012). In America’s federal system, state governments wield power over
a number of important policy areas, so their ability to address policy problems remains an
issue of vital interest. To be sure, a wide variety of factors may influence how much policy
change occurs within a system of governance, including public opinion shifts and preference
significantly associated with policy change.
25

polarization. However, this analysis demonstrates that even seemingly “small” institutional
differences governing legislative agenda-setting have a substantial impact on the amount of
policy change or gridlock a political system encounters. Indeed, when legislative parties are
empowered to set a chamber’s voting calendar and thereby exercise negative agenda control,
significantly less policy change occurs.
By tracing the influence of partisan agenda control through the policymaking process,
this study therefore advances current knowledge not only on legislative institutions but on
aggregate-level policy change as well. By moving beyond roll rates, the results suggest not
just that political parties use agenda-setting rules to benefit their party brand, but that these
institutional features carry with them important ramifications for a political system’s ability
to change policy. Future research should further examine how these and other partisan
institutions alter the ability of a legislature not only to make policy but also to fulfill other
duties such as executive oversight.
26

References
Acemoglu, D. and Johnson, S. (2005). Unbundling institutions.
Journal of Political Economy
, 113(5): 949–995.
Alesina A. and A Drazen. (1991). “Why are stabilizations delayed?”
American Economic Review
81: 1170-1188.
Allison, Paul D., and Richard P. Waterman. (2002). "Fixed–effects negative binomial regression models."
Sociological methodology
32, no. 1: 247-265.
Anzia, S. F., & Jackman, M. C. (2013). “Legislative Organization and the Second Face of Power: Evidence from
US State Legislatures.”
The Journal of Politics
,
75
(01), 210-224.
Bawn, K. (1999). Money and majorities in the Federal Republic of Germany: Evidence for a veto players model of
government spending.
American Journal of Political Science
, 707-736.
Binder, S. A. (1999). “The dynamics of legislative gridlock,” 1947-96.
American Political Science Review
, 519-533.
Binder, S. A. (2003).
Stalemate: Causes and consequences of legislative gridlock
. Brookings Institution Press.
Bowler, Shaun, and Todd Donovan. (2004) "Measuring the effect of direct democracy on state policy: Not all
initiatives are created equal."
State Politics & Policy Quarterly
4, no. 3: 345-363.
Bowling, C. J., & Ferguson, M. R. (2001). Divided government, interest representation, and policy differences:
Competing explanations of gridlock in the fifty states.
Journal of Politics
,
63
(1), 182-206.
Chiou, F. Y., & Rothenberg, L. S. (2003). “When pivotal politics meets partisan politics.”
American Journal of
Political Science
,
47
(3), 503-522.
Chiou, F. Y., & Rothenberg, L. S. (2006). Preferences, parties, and legislative productivity.
American Politics
Research
,
34
(6), 705-731.
Chiou, F. Y., & Rothenberg, L. S. (2008). Comparing legislators and legislatures: The dynamics of legislative
gridlock reconsidered.
Political Analysis
,
16
(2), 197-212.
Chiou, F. Y., & Rothenberg, L. S. (2009). A unified theory of US lawmaking: Preferences, institutions, and party
discipline.
The Journal of Politics
,
71
(4), 1257-1272.
Clinton, J.D., (2012). Using roll call estimates to test models of politics.
Annual Review of Political Science
,
15
,
pp.79-99.
Council of State Governments.
Book of the States
. Lexington, KY: Council of State Governments, 1994-1995, 1996-
1997, 1998-1999, 2000-2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2013.
27

Covington, C. R., & Bargen, A. A. (2004). Comparing Floor‐Dominated and Party‐Dominated Explanations of
Policy Change in the House of Representatives.
Journal of Politics
,
66
(4), 1069-1088.
Cox, G.W., Kousser, T. and McCubbins, M.D., (2010). Party power or preferences? Quasi-experimental evidence
from American state legislatures.
The Journal of Politics
,
72
(3), pp.799-811.
Clinton, J.D. and Lapinski, J.S., (2006). Measuring legislative accomplishment, 1877–1994.
American Journal of
Political Science
,
50
(1), pp.232-249.Cox, G. W., & McCubbins, M. D. (2005).
Setting the agenda:
Responsible party government in the US House of Representatives
. Cambridge University Press.
Diamond, Alexis, and Jasjeet S. Sekhon. (2013). "Genetic matching for estimating causal effects: A general
multivariate matching method for achieving balance in observational studies."
Review of Economics and
Statistics
95, no. 3 :932-945.
Fiorina, M. P. (1996).
Divided government
(2nd ed.). Boston: Allyn and Bacon.
Fong, C., & Krehbiel, K. (2018). Limited Obstruction.
American Political Science Review
,
112
(1), 1-14.
Gailmard, S. and Jenkins, J.A., 2007. Negative agenda control in the Senate and house: Fingerprints of majority
party power.
The Journal of Politics
,
69
(3), pp.689-700.
Gray, V., & Lowery, D. (1995). “Interest representation and democratic gridlock.”
Legislative Studies Quarterly
,
531-552.
Gerber, Elisabeth R. (1996) "Legislative response to the threat of popular initiatives."
American Journal of Political
Science
, 99-128.
Hammond, T.H. and Miller, G.J., (1987). The core of the constitution.
American Political Science Review
,
81
(4),
pp.1155-1174.
Henisz, W. J. (2000). The institutional environment for economic growth.
Economics and Politics
, 12(1): 1–31.
Hicks, W. D., & Smith, D. A. (2009). “Do Parties Matter? Explaining Legislative Productivity in the American
States.” The State of the Parties 2008 & Beyond Conference, The Bliss Institute, University of Akron, Ohio.
Hlavac, Marek (2014). stargazer: LaTeX/HTML code and ASCII text for well-formatted regression and summary
statistics tables. R package version 5.1. http://CRAN.R-project.org/package=stargazer
Howell, W., Adler, S., Cameron, C., & Riemann, C. (2000). “Divided government and the legislative productivity
of Congress, 1945-94.”
Legislative Studies Quarterly
, 285-312.
28

Jackman, M.C., 2013. Parties, median legislators, and agenda setting: How legislative institutions matter.
The
Journal of Politics
,
76
(1), pp.259-272.
Krehbiel, K. (1998).
Pivotal politics: A theory of US lawmaking
. University of Chicago Press.
Kurtz, K. (2007). Filibusters Rare in State Legislatures.
National Conference of State Legislatures.
Retrieved
January, 2018, from http://ncsl.typepad.com/the_thicket/2007/02/filibusters_rar.html.
Lawrence, E. D., Maltzman, F., & Smith, S. S. (2006). Who wins? Party effects in legislative voting.
Legislative
Studies Quarterly
,
31
(1), 33-69.
Mann, Thomas E., and Norman J. Ornstein. (2012).
It's even worse than it looks: How the American constitutional
system collided with the new politics of extremism
. Basic Books.
Mayhew, David R. (1991).
Divided we govern
. Yale University.
National Conference of State Legislatures.
HEALTH REFORM: 2011-2013 STATE LEGISLATIVE TRACKING
DATABASE
. Last modified January 28, 2014. Accessed December 1, 2014.
http://www.ncsl.org/research/health/health-reform-database-2011-2013-state-legislation.aspx.
Reilly, M. (2009).
States Limiting Legislative Debate
(Connecticut, State Assembly, Office of Legislative Research).
Available at https://www.cga.ct.gov/2009/rpt/2009-R-0249.htm.
Richman, J., (2011). Parties, pivots, and policy: The status quo test.
American Political Science Review
,
105
(1),
pp.151-165.
Rogers, J. R. (2005). The impact of divided government on legislative production.
Public Choice
,
123
(1-2), 217-233.
Shor, B., & McCarty, N. (2011). “The ideological mapping of American legislatures.”
American Political Science
Review
,
105
(03), 530-551.
Stiglitz, E. H., & Weingast, B. R. (2010). Agenda control in Congress: Evidence from cutpoint estimates and ideal
point uncertainty.
Legislative Studies Quarterly
,
35
(2), 157-185.
Strickland, James Manning, “Disentangling the Effects of Lobbying Laws on Interest Group Registrations in the
American States, 1988-2013” (July 18, 2014). Available at SSRN: http://ssrn.com/abstract=2467944 or
http://dx.doi.org/10.2139/ssrn.2467944
Tsebelis, G. (2002).
Veto players: How political institutions work
. Princeton University Press.
Tsebelis, G., & Alemán, E. (2005). Presidential conditional agenda setting in Latin America.
World
Politics
,
57
(03), 396-420.
29

Woon, J. and Cook, I.P., 2015. Competing gridlock models and status quo policies. Political Analysis, 23(3),
pp.385-399.
30

Table 1: Bill Enactments and Core Size
Dependent variable:
Enactments Logged Enactments
(1) (2) (3) (4)
Size of Core −0.363∗∗∗
−0.278∗∗∗
−0.172∗
−0.172∗
(0.086) (0.090) (0.087) (0.090)
Bill Introductions 0.00005∗∗∗ 0.00005∗∗∗ 0.00004∗∗∗ 0.00004∗∗∗
(0.00001) (0.00001) (0.00001) (0.00001)
Number of Interest Groups −0.0002∗∗∗
−0.0002 −0.0001 −0.0001
(0.0001) (0.0001) (0.0001) (0.0001)
Professionalism −1.219∗∗
−1.046 −0.183 −0.183
(0.500) (0.648) (0.524) (0.648)
Initiatives 0.020 0.028∗∗
−0.010 −0.010
(0.013) (0.013) (0.015) (0.013)
Partisan Dominance 1.037∗∗ 1.287∗∗ 1.081∗∗ 1.081∗∗
(0.405) (0.524) (0.420) (0.524)
State GDP −0.00000 −0.00000 0.00000 0.00000
(0.00000) (0.00000) (0.00000) (0.00000)
Divided Government 0.016 −0.006 −0.050 −0.050
(0.037) (0.035) (0.034) (0.035)
State Population 0.00000∗∗∗ 0.00000 0.000 0.000
(0.00000) (0.00000) (0.00000) (0.00000)
Average Ideology (Upper Chamber) −0.041 0.023 −0.029 −0.029
(0.087) (0.115) (0.098) (0.115)
Average Ideology (Lower Chamber) −0.034 −0.128 −0.105 −0.105
(0.091) (0.117) (0.098) (0.117)
Number of Legislators −0.0001 0.0002 −0.0004 −0.0004
(0.0005) (0.017) (0.001) (0.017)
Constant 6.558∗∗∗ 6.315 6.308∗∗∗ 6.308
(0.161) (0.383)
Observations 357 357 357 357
R20.247 0.277 0.277
Adjusted R20.220 0.232 0.232
Log Likelihood −2,607.334
θ3.548∗∗∗ (0.256)
Akaike Inf. Crit. 5,240.667
State/year effects? N N Y Y
Panel-corrected standard errors? N Y N Y
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗ p<0.01
31

Table 2: Movement of the ACA Compliance Status Quo
Dependent variable:
ACA Compliance
Size of Core −0.464∗∗
(0.231)
Party of Governor 0.479∗∗∗
(0.177)
Number of Interest Groups −0.0002∗
(0.0001)
Professionalism −3.277∗∗
(1.385)
Partisan Dominance −0.103
(1.528)
State GDP 3.89e-06∗∗
(1.77e-06)
Divided Government −0.128∗
(0.074)
State Population −5.81e-08
(6.82e-08)
Mean Core Ideology −0.144
(0.131)
Constant 2.950∗∗∗
(0.313)
Observations 49
Log pseudolikelihood −158.802
Wald χ2186.43∗∗∗
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗ p<0.01
32

Table 3. Matched Sample
(from GenMatch function)
AK VA
AL ND
AZ LA
CT LA
DE VT
FL CA
GA LA
HI ND
IA OR
IL OR
IN UT
KS SC
KY VA
MA LA
MD LA
MI LA
MN CO
MO NV
MS ME
MT NV
NC AR
NH AR
NJ NV
NM ME
NY CO
OH LA
OK ID
PA CA
RI ND
SD UT
TN UT
TX CA
WA OR
WI UT
WV VT
WY ND
33

34

0.5 1.0 1.5
5
10
15
Size of Core
Predicted value

Supplemental Material
A: Negative Agenda Control and Core Size (H1)
According to H1, the presence of negative-agenda-control institutions should positively pre-
dict core size. I find support for this assertion.
Figure A1 plots the number of additional veto players due to negative agenda control
(ranging from 0 to 2) against core size. As expected, an increase in number of veto players
correlates with an increase in core size. However, as demonstrated by the core size overlap
between the agenda-control categories, absorption is a very real phenomenon in the data, as
are cross-state differences in the overall level of preference polarization.
Figure A1: Negative Agenda Control and Core Size
To examine whether or not this relationship between negative agenda control and core
size is robust to the incorporation of other covariates relevant to core size, I estimate a linear
1

model of core size, first using panel corrected standard errors, and then using state-level ran-
dom effects with year fixed effects. Along with the number of veto players/negative agenda
control, I include the variables Bill Introductions (number of bills introduced during a state-
biennium), Number of Interest Groups,Professionalism,Initiatives (number of initiatives
passed in a state-year), Partisan Dominance (folded six-year Ranney Index), State GDP
(economic output in a state-year, in chained 1997 dollars), a Divided Government dummy,
state-year Population,Average Ideology in both the upper and lower chambers, and total
Number of Legislators in a state. Table A1 summarizes the results.
As Table A1 demonstrates, the presence of negative agenda control does indeed positively
correlate with a state’s core size. This result is significant at the p<.01 level. As predicted,
negative agenda control appears to matter for core size in the aggregate: a larger number
of veto players (via the introduction of negative agenda control) is associated with larger
core sizes. This is not a surprising result, given that the presence of negative agenda control
is used to calculate core size itself. However, it does establish that these data square with
theoretical expectations regarding veto players and core size and offer an opportunity to test
H2 and H3.
2

Table A1: Negative Agenda Control and Core Size
Dependent variable:
Size of Core
(1) (2)
Negative Agenda Control 0.105∗∗∗ 0.100∗∗∗
(0.023) (0.035)
Bill Introductions 0.00002∗∗∗ 0.00001∗
(0.00000) (0.00000)
Number of Interest Groups 0.00004 0.00003
(0.00005) (0.0001)
Professionalism −0.441 0.045
(0.305) (0.306)
Initiatives −0.0001 −0.002
(0.008) (0.009)
Partisan Dominance −1.612∗∗∗
−1.030∗∗∗
(0.232) (0.249)
State GDP 0.00000 0.000
(0.00000) (0.00000)
Divided Government −0.032 −0.013
(0.023) (0.020)
Population −0.000 0.000
(0.00000) (0.00000)
Average Ideology (Upper Chamber) −0.085 −0.220∗∗∗
(0.053) (0.058)
Average Ideology (Lower Chamber) 0.086 0.138∗∗
(0.056) (0.058)
Number of Legislators 0.0003 0.001
(0.0003) (0.001)
Constant 0.447∗∗∗ 0.274∗∗
(0.096) (0.120)
Observations 357 357
R20.136
Adjusted R20.105
Log Likelihood −99.829
Akaike Inf. Crit. 225.658
State effects? N Y
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗ p<0.01
3

B: Policy Change, Core Size, and the Distribution of Status Quos
As I note throughout my analysis, the distribution of status quo policies—and the potential
uneven nature of this distrbution accross states and time—presents a challenge to results
presented in Section III especially. Indeed, if a low amount of policy change is observed,
for example, it could be the result of what the previous legislature accomplished: should
the previous legislature have accomplished many of the potential objectives of the current
legislature, the current legislature is not likely to change policy a great deal. The opposite
holds true if the previous legislature moved policy in a direction opposed that desired by the
current legislature: in that case, an abundance of opportunities for policy change exist. As
noted in Section III, Clinton (2012) and Tsebelis (2002) both note this possibly, and I thank
an anonymous reviewer for underscoring its importance.1
To address this potential challenge, I have introduced two key control variables into
variables found into the models presented in Table 1. The variables are designed to address
the possibility that the availability changeable status quo policies vary, depending on the
actions of the previous legislature. The first variable, Core Shift represents the absolute
value of the difference between the midpoint of the gridlock interval in time t, compared to
time t−1:
Core Shif t =


Corelef tedge +Corerig htedge
2t
−Corelef tedge +Corerig htedge
2t−1



This variable resembles the “alternation” variable empoyed by Tsebelis (2002, chapter
1The importance of this consideration notwithstanding, Clinton (2012) does show that the distribution
of status quo policy may not ultimately influence on average findings regarding gridlock interval size and
policy change.
4

7) and, to a lesser extent, the “Change in Governmental Regime” used by Krehbiel (1998).
Bawn (1999) also employs a similar variable in her analysis. Tsebelis’s alternation variable
represents the distance between the midpoints of the current and previous governments
within a country. Krehbiel’s governmental regime variable delineates whether there was a
change from unfied to divided, divided to unified, or no change in government. The aim of
these variables is to capture how the previous configuration of veto players may influence
the distribution of status quo policies. In Tsebelis’s case, a shift in government preferences
implies that a great deal of opportunities exist for policy change, on the assumption that
the previous government moved policy in an undesireable direction. In Krehbiel’s case, a
switch from divided to unified govnerment may imply that little policy change occurred in
the previous Congress, leaving ample opportunity for such change in the current Congress—
and vice versa.2In the case of this study, the Core Shift variable attempts to measure how
different the previous legislative regime was ideologically, compared the present regime. If,
for example, veto players were primarily conservative in the previous legislature and are
primarily liberal in the current one, the distribution of status quo policies may favor policy
change: the previous legislature may have moved policy in an “undesireable” direction. The
converse applies if veto players in the current legislature are similar to those in the previous
legislature. Thus, the expectation is that Core Shift should be positively associated with
policy change, all else equal.
The second variable, Lagged Enactments, captures a related, though different, potential
correlate with the distribution of status quo policies. Here, the inclusion of a simple lag of the
2It is important to note that, while Krehbiel includes this variable, he hypothesizes that it should not
influence policy change—a finding that is supported in his data.
5

dependent variable measures how much actual policy change occurred prior to the current
legislature. Were the previous legislature successful in changing a great deal of status quo
policies, it may be possible that fewer such status quo policies are available to the current
legislature for changing. On the other hand, it is possible that such productivity pulled
the status quo farther away from the current legislature. Thus, to this variable, I add an
interaction term (Models 4-8) between Core Shift and Lagged Enactments: perhaps Core
Shift only influences the desireability of the status quo, conditional the previous legislature
actually successfully changing it (captured with Lagged Enactments).
Table A2 summarizes the results. As is clear throughout all model specifications, Core
Size remains negatively and statistically significant in its association with policy change.
The new variables encounter more mixed results. Core Shift, for example, exhibits the
hypothesized positive association, although the effect diminishes with the inclusion of state-
and year-level effects. However, Lagged Enactments is consistently and positively associated
with bill enactments. This may suggest that productivity in the previous session moved the
status quo in such a way that created greater opportunities for policy change. Finally, the
interaction term exhibits no significant relationship in any of the models. Taken together,
these results provide evidence that Section III’s results regarding negative-agenda-control
adjusted core size and policy change are robust to possibilty irregularities in the distribution
of status quos.
6

Table A2: Legislative Productivity, Controlling for Legislative History
Dependent variable: Enactments and Logged Enactments
negative OLS OLS OLS negative OLS OLS OLS
binomial binomial
(1) (2) (3) (4) (5) (6) (7) (8)
Core Size −0.187∗∗
−0.203∗∗
−0.202∗∗
−0.203∗∗
−0.189∗∗
−0.205∗∗
−0.200∗∗
−0.201∗∗
(0.082) (0.086) (0.086) (0.086) (0.082) (0.087) (0.086) (0.087)
Lagged Enactments 0.001∗∗∗ 0.594∗∗∗ 0.578∗∗∗ 0.593∗∗∗ 0.001∗∗∗ 0.606∗∗∗ 0.573∗∗∗ 0.585∗∗∗
(0.0001) (0.039) (0.039) (0.039) (0.0001) (0.046) (0.044) (0.044)
Core Shift 0.240∗∗ 0.243∗0.169 0.242∗0.395 1.112 0.102 0.151
(0.117) (0.127) (0.124) (0.289) (0.284) (1.742) (0.289) (0.292)
Lagged Enactments* −0.0002 −0.132 0.00009 0.0001
Size of Core Shift (0.0003) (0.265) (0.0003) (0.0003)
Bill Introductions 0.0001∗∗∗ 0.0001∗∗∗ 0.00005∗∗∗ 0.00005∗∗∗ 0.0001∗∗∗ 0.0001∗∗∗ 0.00005∗∗∗ 0.00005∗∗∗
(0.00001) (0.00001) (0.000009) (0.000009) (0.00001) (0.00001) (0.000009) (0.000009)
Legislative Professionalism −0.708∗
−0.724 −0.856∗
−0.724 −0.735∗
−0.742 −0.848∗
−0.713
(0.429) (0.462) (0.451) (0.462) (0.430) (0.464) (0.453) (0.464)
Initiatives 0.024∗∗ 0.011 0.012 0.011 0.024∗∗ 0.011 0.013 0.012
(0.010) (0.010) (0.451) (0.010) (0.010) (0.010) (0.010) (0.010)
Electoral Competition 0.277 −0.104 −0.122 −0.105 0.272 −0.110 -0.115 −0.096
(0.328) (0.344) (0.346) (0.344) (0.329) (0.345) 0.348 (0.346)
Size of State Economy −0.00000 0.00000 0.0000007 0.0000008 −0.00000 0.00000 0.0000007 0.0000008
(0.00000) (0.00000) (0.0000007) (0.0000007) (0.00000) (0.00000) 0.0000007 0.0000007
Divided Government −0.024 −0.048 −0.045 −0.048 −0.026 −0.050 −0.044 −0.048
(0.030) (0.034) (0.032) (0.034) (0.030) (0.034) (0.032) (0.034)
State Population 0.00000 −0.00000 −1.52e−08 −2.14e−08 0.00000 −0.00000 -1.51e-08 -2.11e-08
(0.00000) (0.00000) (2.37e-08) (2.49e-08) (0.00000) (0.00000) (2.37e-08) (2.50e-08)
Median Senate Ideology 0.061 0.086 0.093 0.086 0.061 0.088 0.093 0.086
(0.076) (0.079) (0.080) (0.080) (0.079) (0.079) (0.080) (0.079)
Median House Ideology −0.058 −0.039 −0.059 −0.039 −0.057 −0.039 −0.059 -0.040
(0.077) (0.080) (0.081) (0.080) (0.077) (0.081) (0.081) (0.081)
Number of Legislators −0.0002 −0.0003 −0.0003 −.0003 −0.0003 −0.0004 −0.0003 -0.0003
(0.0004) (0.0004) (0.0004) (0.0004) (0.0004) (0.0004) (0.0005) (0.0004)
Constant 5.899∗∗∗ 2.470∗∗∗ 2.792∗∗∗ 2.479∗∗∗ 5.895∗∗∗ 2.397∗∗∗ 2.822∗∗∗ 2.522∗∗∗
(0.135) (0.315) (0.283) (0.315) (0.135) (0.348) (0.306) (0.340)
Year Fixed Effects N Y N Y N Y N Y
State Fixed Effects N N Y Y N N Y Y
Observations 338 338 338 338 338 338 338
R20.588 0.565 0.588 0.588 0.565
θ6.099∗∗∗ 6.105∗∗∗
Akaike Inf. Crit. 4,773.734 4,775.393
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗ p<0.01
7

C: Robustness Checks for H3 Analysis
8

Table A7: Modelling Assessment of H3, Using Matching Variables (Enactments)
Dependent variable:
Logged Enactments
Negative Agenda Control −0.237∗∗∗
(0.040)
Distance between Chamber Medians −0.468
(0.385)
House Majority Median to Chamber Median 0.106
(0.274)
Senate Majority Median to Chamber Median 0.239
(0.277)
Distance between Chamber Majority Medians −0.010
(0.212)
House Chamber Median to Senate Majority Median 0.142
(0.337)
Senate Chamber Median to House Majority Median −0.003
(0.275)
Bill Introductions 0.00004∗∗∗
(0.00001)
Number of Interest Groups −0.0001
(0.0001)
Legislative Professionalism −1.222∗∗
(0.518)
Initiatives 0.009
(0.014)
Partisan Dominance 1.581∗∗∗
(0.446)
State GDP −0.00000
(0.00000)
Divided Government 0.023
(0.042)
State Population 0.00000∗
(0.00000)
Average Upper Chamber Ideology 0.042
(0.097)
Average Lower Chamber Ideology −0.146
(0.101)
Number of Legislators 0.0003
(0.001)
Constant 6.416∗∗∗
(0.208)
Observations 358
Year Fixed Effects? Y
R20.336
Adjusted R20.282
Residual Std. Error 0.561 (df = 330)
F Statistic 6.183∗∗∗ (df = 27; 330)
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗ p<0.01
9

Table A8: Modelling Assessment of H3, Using Matching Variables (ACA Implementation)
Dependent variable:
ACA Implementation
Chambers with Negative Agenda Control −0.359∗∗∗
(0.079)
Distance between Chamber Medians −2.222∗∗
(1.069)
House Chamber Median to Majority Median −1.999∗∗∗
(0.630)
Senate Chamber Median to Majority Median −1.584∗∗∗
(0.588)
Distance Between Majority Medians −3.034∗∗∗
(0.668)
Senate Majority Median to House Chamber Median 2.869∗∗∗
(0.883)
Senate Chamber Median to House Majority Median 3.446∗∗∗
(0.806)
Democratic Governor 0.258∗
(0.157)
Number of Interest Groups −0.0001
(0.0001)
Legislative Professionalism −3.420∗∗∗
(1.002)
2008 Obama Vote Share 0.973
(1.028)
Partisan Dominance 0.524
(0.971)
State GDP 0.00000∗∗∗
(0.00000)
Divided Government −0.159∗∗
(0.076)
Average Core Ideology −0.133
(0.150)
Constant 2.614∗∗∗
(0.598)
Observations 49
Log Likelihood −147.296
θ14.349∗(7.325)
Akaike Inf. Crit. 326.591
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗ p<0.01 ; model is Negative Binomial
10

Figure A2: Marginal Effects for Modelling Assessment of H3 Using Matching Variables -
Enactments
0.0 0.5 1.0 1.5 2.0
6.2
6.4
6.6
6.8
Chambers with Negative Agenda Control
Predicted value
Figure A3: Marginal Effects for Modelling Assessment of H3 Using Matching Variables -
ACA Compliance
0.0 0.5 1.0 1.5 2.0
10
15
20
Chambers with Negative Agenda Control
Predicted value
11

Table A9: Decomposition of Core Analysis (Enactments)
Dependent variable:
Logged Enactments
(1) (2) (3) (4)
Core Size - Chamber Distance −0.588∗∗∗
−0.610∗∗∗
(0.160) (0.161)
Distance between Chamber Medians −0.228∗
−0.129 −0.233∗
−0.132
(0.126) (0.127) (0.126) (0.127)
Bill Introductions 0.00004∗∗∗ 0.0001∗∗∗ 0.00004∗∗∗ 0.0001∗∗∗
(0.00001) (0.00001) (0.00001) (0.00001)
Legislative Professionalism −0.940∗
−1.149∗∗
−0.939∗
−1.143∗∗
(0.502) (0.497) (0.520) (0.514)
Initiatives 0.037∗∗∗ 0.035∗∗∗ 0.038∗∗∗ 0.036∗∗∗
(0.013) (0.013) (0.013) (0.013)
Partisan Dominance 1.262∗∗∗ 1.055∗∗∗ 1.269∗∗∗ 1.061∗∗∗
(0.405) (0.403) (0.408) (0.405)
State GDP −0.00000 −0.00000 −0.00000 −0.00000
(0.00000) (0.00000) (0.00000) (0.00000)
Divided Government 0.015 0.014 0.006 0.006
(0.038) (0.038) (0.041) (0.040)
State Population 0.00000∗∗ 0.00000∗∗ 0.00000∗∗ 0.00000∗∗
(0.00000) (0.00000) (0.00000) (0.00000)
Average Ideology - Upper Chamber 0.023 0.027 0.012 0.014
(0.088) (0.087) (0.089) (0.088)
Average Ideology - Lower Chamber −0.136 −0.133 −0.125 −0.119
(0.094) (0.092) (0.095) (0.093)
Number of Legislators −0.0003 0.00000 −0.0003 0.00000
(0.001) (0.001) (0.001) (0.001)
Constant 6.184∗∗∗ 6.253∗∗∗ 6.193∗∗∗ 6.269∗∗∗
(0.158) (0.157) (0.259) (0.256)
Observations 419 419 419 419
R20.214 0.239 0.228 0.255
Adjusted R20.193 0.217 0.189 0.215
Residual Std. Error 0.592 (df = 407) 0.583 (df = 406) 0.594 (df = 398) 0.584 (df = 397)
F Statistic 10.062∗∗∗ (df = 11; 407) 10.626∗∗∗ (df = 12; 406) 5.867∗∗∗ (df = 20; 398) 6.454∗∗∗ (df = 21; 397)
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗ p<0.01
12

Table A10: Decomposition of Core Analysis (ACA Compliance)
Dependent variable:
ACA Implementation Counts
(1) (2)
Distance between Chamber Medians −0.297 −0.096
(0.271) (0.243)
Core Size - Chamber Distance −1.292∗∗
(0.598)
Democratic Governor 0.449∗∗ 0.470∗∗∗
(0.185) (0.162)
Legislative Professionialism −3.184∗∗
−2.564∗∗
(1.322) (1.223)
Partisan Dominance 0.243 0.304
(1.471) (1.379)
State GDP∗0.00000 0.00000∗
(0.00000) (0.00000)
Divided Government −0.109∗∗∗
−0.107∗
(0.071) (0.062)
Number of Interest Groups −0.0003∗∗∗
−0.0002∗
(0.0001) (0.0001)
State Population −0.00000 −0.00000
(0.00000) (0.00000)
Average Core Ideology −0.119 −0.171
(0.169) (0.158)
Number of Legislators −0.0004 0.0004
(0.001) (0.001)
Constant 2.881∗∗∗ 2.767∗∗∗
(0.267) (0.267)
Observations 49 49
Log Likelihood −161.063 −158.372
θ4.500∗∗∗ (1.292) 5.201∗∗∗ (1.556)
Akaike Inf. Crit. 344.125 340.745
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗ p<0.01
13

D: Balance Statistics for Matching Analysis
Table A11
14

E: Controlling for Possibility of Filibuster
While very little literature exists on filibustering activity in state legislatures, at least
two accounts of debate length rules exist: a 2009 report compiled by a researcher with
the Connecticut state legislature (https://www.cga.ct.gov/2009/rpt/2009-R-0249.htm,
hereafter, the “Connecticut Report”) and a 2007 report by two researchers with the National
Conference of State Legislatures. These sources provide useful information on possible su-
permajoritarian practices in state legislatures, but including them in the analysis implies
some key ambiguities and difficulties. I detail these difficulties below and demonstrate that
the main results concerning core size and policy change remain unchanged.
According to both the Connecticut and NCSL reports, a filibuster is impossible in the
vast majority of states across the U.S. (36 in total), as most chambers possess strict, written
limits on the length of time any legislature or debate may last. In the remaining 14 states, a
filibuster is possible, indicating that filibuster activity could drive policy change downward in
those states. However, this connection between policy change and the presence and absence
of such rules is ambiguous: just because filibustering is possible does not guarantee that
filibustering has ever occurred in the chamber in question (Kurtz 2007). Further complicating
the measurement of a filibuster pivot is the presence or absence of cloture rules. That is,
in at least four of the cases for which a filibuster is possible, no cloture rule appears to
exist, according to the Connecticut report on debate limitations.3Finally, in some states,
there is ambiguity on what kind of vote is needed to invoke cloture, even when such a
rule exists. Indeed, if the cloture requirement is a simple majority, then filibustering may
not add to policy change. However, as Fong and Krehbiel (2018) underscore, even small
3In the extreme, this could suggest that every legislator is a veto player.
15

inconveniences or acts of obstruction can have dramatic influence on policymaking. Taken
together, these difficulties and ambiguities constitute the primary reasons for not formally
including filibuster pivots in my gridlock interval calculations: whether or not a filibuster
pivot exists in a state is itself ambiguous, and even if a filibuster is possible, cloture rules
(which are necessary for locating the filibuster pivot) are not universally clear or available.
In spite of these difficulties, I introduce three variables to control for possibility that
the presence of a filibuster pivot slows policy change. The first variable, drawn from the
Connecticut Report, is a three-way categorical variable that accounts for whether or not a
chamber has predetermined limits on speech. The variable takes on the value “2” if both
chambers lack a speech limit, “1” if only one chamber lacks such limits, and “0” if neither
chamber lacks such limits. Thus, higher values of this variable, which I call Filibuster Pos-
sible, indicate a higher probability that filibustering occurs within a state. I expect this
(and all filibuster-related variables) to be negatively associated with policy change. The
second variable, Filibuster Rule, indicates whether or not a state possesses explicit cloture
rules governing “unlimited” speech. This variable is drawn from a combination of both the
Connecticut and NCSL reports and takes on a “2” if both of a state’s chambers possess a
cloture rule, “1” if just one chamber possesses such a rule, and “0” otherwise. Because clo-
ture rules imply that debate limitations are necessary (i.e., that filibusters actually occur),
higher values of this variable should indicate a larger probability of filibustering. Finally, the
third variable I employ, Cloture Threshold, incorporates supermajoritarian cloture thresholds
more directly. For this variable, I first gather available measures on supermajoritarian cloture
thresholds, found within the NCSL reports. Then, for the states in which supermajoritarian
cloture requirements exist, Cloture Threshold takes on the value of that requirement—and
16

zero otherwise. Thus, for a state with a Senate requiring three-fifths majority to invoke clo-
ture, Cloture Threshold=.6. If both chambers possess supermajority requirements, Cloture
Threshold takes on the value of the sum of supermajorities. Thus, in a state where both
chambers require three-fifths majority to end debate, Cloture Threshold = 1.2. Assuming
that higher majority thresholds strengthen the power of a filibuster, I again expect that
higher values of this variable should be negatively associated with policy change.
Table A12 summarizes the results. Models 1-3 are negative binomial models with no
state- or year-level effects, models 4-6 are OLS models of logged enactments with year-level
effects, and models 7-9 are OLS models of logged enactments with both year- and state-level
effects. In all cases, the presence of any filibuster-related variable strengthens the statistical
relationship between core size and bill enactments, which remains negative and statistically
significant across all model specifications. The filibuster variables themselves, however, ex-
hibit less consistency. In models not including state-level effects, all three filibuster variables
exhibit a negative and statistically signifcant relationship with policy change, as expected.
However, when state-level effects are introduced, the variables fall out of significance. Nev-
ertheless, the filibuster variables appear to function largely as expected.
Taken together, these results indicate that, while filibuster rules may play a part in
determining how much policy change occurs within a state, they do not serve as a confounder
for the observed relationship between the agenda-control-adjusted core variable and policy
change. Future research may consider further exploration of debate rules in state legislatures,
in order to determine if and where actual filibuster pivots exist—thereby better enabling their
inclusion in gridlock interval calculations.
17

Table A12: Filibuster Pivots and Policy Change
Dependent variable:
Count of Enactments Logged Enactments
negative OLS
binomial
(1) (2) (3) (4) (5) (6) (7) (8)
Size of Core −0.391∗∗∗
−0.389∗∗∗
−0.323∗∗∗
−0.341∗∗∗
−0.333∗∗∗
−0.257∗∗∗
−0.167∗∗
−0.167∗∗
−0.150∗∗
(0.087) (0.086) (0.089) (0.096) (0.096) (0.098) (0.072) (0.072) (0.074)
Cloture Threshold −0.175∗∗
−0.173∗∗
−0.155
(0.069) (0.077) (0.154)
Cloture Rule −0.175∗∗∗
−0.145∗∗
−0.134
(0.066) (0.073) (0.137)
Filibuster Possible −0.103∗
−0.103 −0.106
(0.057) (0.064) (0.117)
Bill Introductions 0.00005∗∗∗ 0.00005∗∗∗ 0.00005∗∗∗ 0.00005∗∗∗ 0.00005∗∗∗ 0.00004∗∗∗ 0.00004∗∗∗ .00004∗∗∗ 0.00004∗∗∗
(0.00001) (0.00001) (0.00001) (0.00001) (0.00001) (0.00001) (0.00001) (0.00001) (0.00001)
Number of Interest Groups −0.0002∗∗∗
−0.0002∗∗∗
−0.0002∗∗∗
−0.0002∗∗
−0.0002∗∗
−0.0002∗∗
−0.0001 -0.0001 -0.0001
(0.0001) (0.0001) (0.0001) (0.0001) (0.0001) (0.0001) (0.00008) (0.00008) (0.00008)
Professionalism −1.327∗∗∗
−1.423∗∗∗
−1.059∗∗
−1.246∗∗
−1.294∗∗
−0.897 −0.134 -0.154 -0.054
(0.499) (0.503) (0.508) (0.574) (0.579) (0.586) (0.624) (0.622) 0.646
Initiatives 0.015 0.015 0.022∗0.021 0.022 0.027∗
−0.015 -0.015 -0.014
(0.013) (0.013) (0.013) (0.014) (0.014) (0.014) (0.010) (0.010) (0.010)
Electoral Competition 1.187∗∗∗ 1.175∗∗∗ 1.024∗∗ 1.362∗∗∗ 1.366∗∗∗ 1.232∗∗∗ 1.080∗1.085∗1.016∗
(0.403) (0.403) (0.404) (0.447) (0.448) (0.448) (0.564) (0.562) (0.557)
Size of State Economy −0.00000 −0.00000 −0.00000 −0.00000 −0.00000 −0.00000 7.14e-07 6.74e-07 7.45e-07
(0.00000) (0.00000) (0.00000) (0.00000) (0.00000) (0.00000) (7.25e-07) (7.34e-07) (7.23e-07)
Divided Government 0.015 0.019 0.016 −0.006 −0.002 −0.007 −0.052 −0.051 -0.053
(0.037) (0.037) (0.037) (0.043) (0.043) (0.043) (0.038) (0.038) (0.038)
State Population 0.00000∗∗∗ 0.00000∗∗∗ 0.00000∗∗∗ 0.00000∗∗ 0.00000∗∗ 0.00000∗∗ 2.51e-09 4.69e-09 1.98e-09
(0.00000) (0.00000) (0.00000) (0.00000) (0.00000) (0.00000) (2.89e-08) (2.93e-08) (2.82e-08)
Average Senate Ideology −0.040 −0.046 −0.052 0.003 0.001 −0.013 -0.039 -0.038 -0.044
(0.087) (0.087) (0.088) (0.096) (0.096) (0.097) (0.071) (0.071) (0.072)
Average House Ideology −0.057 −0.071 −0.014 −0.138 −0.141 −0.092 −0.113 -0.114 -0.102
(0.091) (0.091) (0.092) (0.101) (0.101) (0.102) (0.070) (0.070) (0.070)
Number of Legislators −0.0002 −0.0003 −0.0002 −0.0001 −0.0002 −0.0001 - 0.0006 -0.0006 -0.0006
(0.0005) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001)
Constant 6.654∗∗∗ 6.679∗∗∗ 6.579∗∗∗ 6.580∗∗∗ 6.571∗∗∗ 6.482∗∗∗ 6.369∗∗∗ 6.372∗∗∗ 6.350∗∗∗
(0.166) (0.168) (0.161) (0.463) (0.464) (0.463) (0.322) (0.328) (0.310)
State Fixed Effects N N N N N N Y Y Y
Year Fixed Effects N N N Y Y Y Y Y Y
Observations 357 357 357 357 357 357 357 357 357
R20.280 0.278 0.275 0.232 0.231 0.233
θ3.606∗∗∗ 3.611∗∗∗ 3.575∗∗∗
Akaike Inf. Crit. 5,236.436 5,235.896 5,239.685
Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗ p<0.01
18

F: Search Interface for Calculating ACA Measure
Figure A4: Constructing the ACA Implementation Measure
19