Preferences, Parties, and Legislative Productivity

• Author: Lawrence S. Rothenberg,Fang-Yi Chiou
• DOI: 10.1177/1532673X06289159
• Published date: 01 November 2006
• Date: 01 November 2006
• Subject Matter: Articles

American Politics Research
Volume 34 Number 6
November 2006 705-731
© 2006 Sage Publications
10.1177/1532673X06289159
Preferences, Parties, and
http://apr.sagepub.com
hosted at
Legislative Productivity
http://online.sagepub.com
Fang-Yi Chiou
Academia Sinica, Taipei, Taiwan
Lawrence S. Rothenberg
University of Rochester, New York
Chiou and Rothenberg (2003) utilized models of legislative productivity,
synthesizing parties and preferences by deriving and measuring different
equilibrium gridlock intervals, finding that such models outperformed alter-
natives depending on preferences alone. However, their temporal scope is
limited to post-World War II. Generalizing theoretical models—notably rec-
ognizing the 1917 enactment of Rule 22 creating the Senate filibuster rule—
and expanding historical scope empirically will produce results, inspiring
greater confidence and better ability to distinguish whether partisan features
matter. Hence, the authors generalize and test models using data extending to
the late 19th century. Results indicate that a model combining preferences
with party unity better explains outcomes than does a preference-based
model, a model where parties control the agenda but cannot discipline
members, or a model where the president rules his legislative party. In addi-
tion, findings are consistent with the inference that Rule 22 did not change
the filibuster rule guiding the Senate substantially.
Keywords:
legislative majorities; U.S. Congress; filibusters; legislative
gridlock; political party unity
In recent years, scholars have offered competing perspectives and gener-
ated differing empirical findings on legislative productivity—the extent
of significant legislative output—and gridlock. One explanation for such
contrasts is difficulties in linking theory and data, as competing theories
Authors’ Note: Thanks to Larry Bartels, Patrick Brandt, Joshua Clinton, Bo Honoré, and
Brian Sala for advice and, especially, to Keith Poole for extending presidential common space
score measurement. Technical appendices are available from the authors and will be posted on
the Web.
705

---

706
American Politics Research
must be clearly defined and tests require appropriate data. More specifi-
cally, two conditions must be satisfied. First, preferences and parties must
be synthesized into testable theories. This has proven problematic as theo-
rists, most notably Krehbiel (1996, 1998; see also Brady & Volden, 1998),
have rejected the relevance of parties, and empirical party scholars have not
theoretically linked their work to capture the potentially complicated and
dynamic relationship between parties and preferences (e.g., Binder, 2003).
Second, data must be of sufficient quality and quantity to produce confi-
dence in empirical results. Although earlier work by Chiou and Rothenberg
(2003) may meet the first criterion, previous analyses may be criticized as
not satisfying the second, particularly by estimating models using data over
a brief historical era.
Nor does collecting data during a longer period simply solve this prob-
lem, as temporal changes in institutional rules complicate theoretical test-
ing. Most important for gridlock models, the cloture requirement created by
Rule 22’s 1917 passage in the U.S. Senate poses a challenge. Specifically,
filibuster pivots are a key component for measuring equilibrium gridlock
intervals (EGIs)—which are the set of status quo policies at the beginning
of a Congress that, in equilibrium, are unchanged when that Congress
concludes—in pivotal politics theories and their partisan extensions.
With this in mind, we generalize Chiou and Rothenberg’s (2003) grid-
lock models by accounting for changes in the filibuster rule and generate
EGIs capturing the impacts of preferences and parties that are amenable to
historical empirical testing. We then test these models using data extending
back to the late 19th century to allow us to draw inferences confidently
about what role parties and their leaders play, if any, relative to actor pref-
erences for the EGIs realized and the legislative choices made. Our findings
not only provide a key piece of evidence in distinguishing between whether
preference-based approaches or those integrating parties best capture leg-
islative decision making but also allow us to distinguish which partisan fea-
tures appear to matter. In addition, our research provides insight into the
importance of incorporating the filibuster rule for estimating the gridlock
interval and the impact of the adoption of Rule 22.
Thinking About the Pre-1917 Filibuster Rule
Although our principal concern involves understanding the role of pref-
erences and parties for legislative productivity, given its centrality for deter-
mining the EGIs, the pre-1917 filibuster requires some commentary (on

---

Chiou, Rothenberg / Legislative Productivity
707
Rule 22, see Bawn & Koger, 2003; Binder & Smith, 1997; Koger, 2002;
Mayhew, 2003; Schickler & Wawro, 2003; Wawro & Schickler, 2003).
Although there was a formal rule pre-Rule 22—each member having the
right to filibuster and unanimity required for proceeding—assuming una-
nimity has little plausibility and would not produce a model with meaning-
ful predictive capacity.
Indeed, by all accounts, the founding fathers considered the Senate a
majoritarian institution that would be slowed by member deliberation rather
than rules (Binder & Smith, 1997). Although this presumption appeared to
break down somewhat in the late 19th century, there is reason to believe
that Rule 22 de facto expanded or ratified the supermajority required for
legislative passage.
Besides historical observation, there are several reasons why assuming
unanimity is unreasonable. For one thing, late 19th century Senate prece-
dents mitigated the effectiveness of dilatory motions (Burdette, 1940). For
another, indefinite filibustering by any member or group of members is
very costly (e.g., Beth & Bach, 2003; Wawro & Schickler, 2003). Also,
because legislators are interacting so many times on different legislation,
they may be able to generate a norm that establishes a de facto filibuster
rule well below 100%; indeed, the breakdown of such a tacit agreement
might have necessitated the formalization embodied in Rule 22.1
Furthermore, empirical analysis suggests that the filibuster pivot was no
more than two thirds in earlier years. For example, Wawro and Schickler
(2003) maintain that the empirical evidence indicates the 1917 changes
induced larger supermajorities.2 Although existing evidence is not defini-
tive given the possibility of other factors working and the need for a more
precise theory of coalition size and filibuster pivots, it is consistent with the
cut point being upper bounded at two thirds.3
Given the near consensus that the filibuster requirement was, in general,
either strengthened or maintained in 1917, we assume that Rule 22 either
ratified a shaky status quo or somewhat raised the filibuster requirement. As
this leaves no magical formula for setting the filibuster pivot, we principally
distinguish among: (a) pure majoritarianism, which almost certainly did not
characterize the situation pre-Rule 22; (b) modest supermajority—as an
intuitive range to settle on if Rule 22 strengthened the filibuster rule, a vari-
ety of points between pure majoritarianism and 67%; and (c) post-1917
supermajority (67%), capturing the possibility that Rule 22 formalized the
status quo. However, for generality and to capture what a narrow reading of
the Senate rules would produce, we also refer to findings where the fili-
buster is assumed to be beyond 67%.

---

708
American Politics Research
Competing Theories of Gridlock: Review
and Generalization
We now review and generalize Chiou and Rothenberg’s (2003) four
competing models, sketching key characteristics and determinants of the
EGIs for each and indicating how the lack of a pre-1917 filibuster rule
affects expectations by focusing on how varying the cloture requirement
changes the gridlock interval. Our discussion is nontechnical and highlights
how different EGIs may be, and are, produced (not all possible EGI types
are realized in our data).
All four models have common assumptions—unidimensional policy space
(e.g., liberalism-conservatism), single-peaked preferences, bicameralism, and
the existence of left and right filibuster pivots—but differ in how parties are
conceptualized.4 In one model parties have no role, in another parties function
as agenda-setters, in a third they also control their members, and in a fourth
presidents also serve as the leader for their party’s members.
Basic Model (Preferences With No Party Role)
This model is similar to Krehbiel’s (1998) pivotal politics model except
that it includes a second legislative chamber and a second filibuster pivot
closer to the president. There are five players—the House and Senate
median voters and override pivots, the liberal and conservative filibuster
pivots in the Senate, and the president—and a game sequence: (a) the
Senate median proposes a bill; (b) each filibuster pivot chooses whether to
filibuster (maintaining the status quo) or to let the game proceed to the next
stage; (c) the House median rejects the bill (again maintaining the status
quo) or accepts it to allow the game’s continuation; (d) the president signs
or vetoes the bill; (e) given a veto, the chamber override pivots decide either
sustain or override, with the status quo only changing if both override.
When this game is solved with different cloture requirements, we find
that increasing cloture...