1
How Do Committee Assignments Facilitate Legislative
Party Power? Evidence from a Randomized Experiment
in the Arkansas State Legislature*
David E. Broockman
Graduate Student
Department of Political Science
University of California, Berkeley
210 Barrows Hall
Berkeley, CA 94720
[email protected]
Daniel M. Butler
Assistant Professor
Department of Political Science
Yale University
P.O. Box 208209
New Haven, CT 06520-8209
[email protected]
Abstract
How does the committee assignment process facilitate legislative party power? This paper sheds
light on this longstanding question by exploiting unique rules in Arkansas’ state legislature,
where legislators select their own committee assignments in a randomized order. The natural
experiment reveals that the legislators reap at most limited rewards from winning their preferred
assignments. These results are consistent across time and legislators’ careers. These results
suggest that committee seats have at best limited use as inducements for party leaders seeking to
incentivize party loyalty. Instead, party leaders’ careful attention to their members’ loyalty
during the committee assignment process in most legislators may simply be explained by party
leaders’ desire to control committees by stacking them with loyal legislators, not a quest to
generate incentives for legislators to remain loyal.
*
The authors’ names appear in alphabetical order and both contributed equally to this paper. We thank seminar
participants at Yale, the 2012 meeting of the Midwest Political Science Association, and the 2011 State Politics and
Policy Conference as well as Jay Barth, Jim Battista, Keith Krehbiel, Lynda Powell, Eric Schickler, and Rob van
Houweling for helpful comments. Replication materials for this study will be posted at Yale’s Institution for Social
and Policy Studies data archive (http://isps.research.yale.edu). Funding and institutional support were provided by
Yale’s Institution for Social and Policy Studies. David Broockman also acknowledges the National Science
Foundation Graduate Research Fellowship Program for support. The authors thank Roger Potts, the Arkansas
Legislative Digest, and many staffers at the Arkansas state legislature for their help with acquiring this data.
Cameron Rotblat and Stacey Chen also provided excellent research assistance.
2
Why do American legislators vote so frequently with their copartisans (e.g., Fowler and
Hall 2013)? Chief among legislative scholars’ many explanations for legislators’ party unity has
been the view that party leaders use their control over the committee assignment process to
incentivize their members to vote with the party. The logic implicating the committee assignment
process in party loyalty is straightforward. Seats on powerful committees are widely thought to
represent a “considerable asset” for legislators (e.g., Fenno 1973; Grimmer and Powell 2013).
Party leaders generally control assignments to such committees. Therefore, party leaders are
thought to use desirable assignments as significant carrots (and sticks) to incentivize their
members. Consistent with this view, party leaders have been found to consistently assign their
most loyal members to the most powerful committees; and, legislators assigned to such
committees are also typically more successful at being re-elected, passing new policy, and
securing influence in their chamber.
We present novel evidence that tests the micro-foundations of this conventional wisdom
about the importance of committee assignments in generating party unity – namely, whether
legislators meaningfully benefit from serving on their preferred committees. Our evidence comes
from the Arkansas state legislature, where parties play no role in making standing committee
assignments; instead, legislators choose their own assignments in the order of their seniority.
Crucially, for legislators who have served the same length of time, this seniority order is
determined by a random lottery. Some members are thus randomly assigned to have a better
opportunity to select their preferred assignments, a situation essentially equivalent to randomly
assigning party leaders’ intention to reward some members with access to their most preferred
committee assignments and punish others with the dregs.
We use this randomized lottery to test whether legislators who have a more complete and
higher quality slate of committee assignments to choose from gain appreciable benefits over
3
those who are forced to accept the assignments no other legislators want. If winning preferred
committee seats is so valuable to legislators that the quest for valuable committee seats could
form the basis for legislative party power, we would expect such legislators to be considerably
more likely to achieve their electoral, policy, and political goals as a result of gaining their
preferred committee assignments. However, in contrast to a great deal of previous observational
work, the results from the randomized lottery suggest that legislators are not significantly more
likely to attain any of their principal goals as a result of attaining their preferred assignments.
This evidence suggests that the benefits individual legislators derive from having better
committee assignments is too meager to form a plausible basis for party power.
In concluding, we suggest that party leaders may use their control over the committee
assignment process primarily to stack powerful committees with loyalists, not to induce party
loyalty with the promise of plum assignments. Because winning preferred committee seats
appears to offer at most meager benefits to legislators, such assignments are unlikely to be the
potent instruments for sustaining party unity that scholars often portray them. Rather, our
findings support the view that scholars should look beyond the committee assignment process to
understand legislative party unity and beyond the desire to promote party unity to understand
committee assignments.
The Theoretical Micro-foundations of Committee Assignment-Induced Legislative Party
Unity
The theoretical foundation of the committee assignment-based theory of legislative party
unity is scholars’ view that legislators gain appreciable advantages from serving on influential
committees. Fenno (1973), for example, first identified legislators’ three main goals – achieving
re-election, crafting good public policy, and securing prominence in their chamber – and argued
4
that committee memberships helped legislators achieve all of them. Scholars since Fenno have
consistently supported this view that winning preferred committee assignments helps legislators
achieve their central goals. Most commonly, scholars find that legislators who receive their
preferred committee assignments are more likely to win re-election (e.g., Mayhew 1974; Bullock
1976; Grimmer and Powell 2013 among many).1
As Shepsle (1978, p. 35) summarizes, “the seeking and obtaining of desirable committee
assignments” is thought to be “a matter of the highest priority” for legislators. Preferred
assignments are theorized to help legislators win re-election for three main reasons. First,
legislators tend to sit on committees that concern policy issues relevant to their districts and can
strongly influence policy on issues that are particularly important to their constituents, allowing
them to claim credit for policy changes on issues that are particularly salient to their constituents
(Fenno 1973; Mayhew 1974; Weingast and Marshall 1988; Adler and Lapinski 1997). Second,
seats on certain committees may also help legislators steer more government money to their
districts (Carsey and Rundquist 1999; Stewart and Groseclose 1999). Finally, interest groups
give more money to legislators assigned to committees with greater influence, especially over
areas relevant to them (Grier and Munger 1991, 1993; Romer and Snyder 1994; Dow, Endersby,
and Menifield 1998). Scholars also argue that certain committee assignments help legislators
gain prestige among colleagues as they vie for leadership positions (Manley 1970; Shepsle 1978;
Fenno 1973) and to help them craft good public policy (Fenno 1973).
Existing Literature: Controlling Committee Assignments Allows Legislative Parties To
Incentivize Party Loyalty
Building on the view that winning preferred committee assignments provides legislators
1
See also Shepsle (1978), Fowler, Douglass, and Clark (1980), Smith and Deering (1983); Katz and Sala (1996),
Crain and Sullivan (1997), Milyo (1997), Maltzman (1997), Leighton and Lopez (2002), and Heberlig (2003).
5
with considerable benefits, scholars have subsequently argued that party leaders’ control over the
committee assignment process represents a (if not the) central tool for them to incentivize party
unity among rank-and-file members. Committee assignments are one of the few excludable
benefits party leaders appear to have at their disposal, and thus it is quite reasonable to implicate
them in these leaders’ quest to influence their members given their apparent attendant rewards
(see, among many, Rohde and Shepsle 1973; Westefield 1974; Crook and Hibbing 1985; Coker
and Crain 1994; Sinclair 1995, p. 93; Stratmann 2000, p. 666; Snyder and Groseclose 2000, p.
194; Roberts and Smith 2003; Kanthak 2004). As Smith (2000, p. 62) reviews scholars’ enduring
view, “tangible incentives [parties use to enforce loyalty] come in many other forms, although
few are as important as committee assignments to most legislators.” Likewise, Cox and
McCubbins (1993, p. 175) find that “loyalty to the party leadership is a statistically and
substantively important determinant of who gets what [committee] assignment”, leading them to
conclude that the committee assignment process makes legislators “more responsive to both the
party’s leadership and goals” (p. 182).
The view that partisan control of the committee assignment process is central to
legislative parties’ ability to incentivize their members has heretofore met little skepticism. Only
one study of which we are aware – Bullock (1972) – has produced evidence that more desirable
assignments do not help freshmen legislators attain their re-election goal.2 Likewise, Krehbiel
(1993) stands apart from the literature by raising doubts that parties play a role in causally
influencing legislators’ voting behavior, and we know of no studies that have cast doubt on the
proposition that parties can use committee assignments as significant inducements.
2
On a related though distinct point, Kellerman and Shepsle (2009) find that Members of Congress with less
seniority on a committee are less likely to serve as committee chair and have fewer sponsored bills passed in the
jurisdiction of their committee as a result; however, they find no effect of this increased likelihood of service as a
committee chair on legislators’ re-election prospects.
6
Endogeneity in the Party-Led Assignment Process
Despite the theoretical significance of committee seats’ value to legislators, estimating
the benefits members accrue from winning their preferred committee assignments is difficult
because we do not fully understand why legislators are assigned to the committees that they are.
Without full knowledge of the criteria party leaders use to make assignments or why legislators
request certain committees, we cannot adequately control for the factors that might influence
their assignments in order to distinguish the effects of assignment to certain committees from
pre-existing differences in their members’ characteristics that make them more likely to be
assigned to them. If members’ expected success in winning re-election, fundraising, or gaining
influence correlates with the reasons different legislators end up on different committees,
comparing how members of different committees behave could thus mistake cause and effect.
Indeed, a number of studies have concluded that such factors do influence legislators’ committee
choices and party leaders’ committee assignments, leaving existing studies vulnerable to
potentially severe omitted variable bias (see Masters 1961; Shepsle 1978; Hedlund and Patterson
1992; Hedlund et al. 2011).
In the next section we discuss the natural experiment in the Arkansas state legislature,
where a randomized lottery helps determine legislators’ committee assignments instead of party
leaders.
Research Design: The Randomized Committee Lottery in Arkansas
Each legislator in Arkansas’ two state legislative chambers has a seniority number, and
legislators choose their own standing committee assignments in the order of this seniority
7
number.3 This seniority number is first determined by how long a member has served in the
chamber, with the lowest numbers (and thus the first choice of committee assignments) going to
those who have served longest. Crucially, however, the seniority number of legislators who have
served the same length of time is randomly determined: before their first term, legislators draw
numbers written on slips of paper out of a hat to determine their seniority within their freshman
class.4 Their relative seniority within their cohort stays with them for the remainder of their time
in the legislature.
The Independent Variable: Relative Rank
There is one slight complication in the Arkansas committee assignment lottery: although
legislators’ seniority is randomized across their entire cohort, committee assignments in the
House are allotted within four separate ‘caucus districts’ corresponding to the four congressional
districts in Arkansas. Because only a certain pre-set number of legislators from each caucus
district can sit on each committee, House members only compete with other legislators within
their caucus district for committee seats.
Table 1 illustrates how we operationalize the full randomization that occurs within these
chamber-cohort-caucus district groupings for estimating the effects of the quality of legislators’
committee assignments. In the Table we present a fictional 25-member Arkansas House
populated with legislators in their first or second term. Legislators in their first term have lower
seniority numbers than the legislators in their second term. However, because seniority within
each cohort is randomly determined, what matters for the natural experiment is one’s seniority
3
Legislators serve on two standing committees. Legislators choose their first committee in the order of seniority and
then choose their second committee in the same order. As Figure 1 shows and we discuss, this arrangement leads to
significant heterogeneity in the quality of legislators’ committee assignments, with legislators choosing first
systematically serving on different committees who choose later in the process.
4
Term lengths following decennial elections are also randomized in Arkansas (Gaines, Nokken, and Groebe 2012).
8
rank within one’s chamber-cohort-caucus district group. The fictional legislator assigned the
seniority number 4 in Table 1 is therefore actually in a more advantageous position than the
legislator assigned a 3, because legislator 4 picks first in his caucus district, whereas legislator 3
picks only after legislator 2 because they are both in caucus district B.
[INSERT TABLE 1 HERE]
The resulting Relative Rank metric is shown for our fictional legislature in part (b) of
Table 1. Legislators are arranged in groups by their cohort and caucus district and then sorted by
their randomized seniority number within these groups because legislators pick their own
assignments in direct succession within these groups. Thus, for example, within caucus district
A, legislator 4 would pick first, followed by legislator 6 and 11. Once the senior members finish
picking, legislators 14, 15, 18, and 23 would then pick the remainder of the assignments
allocated to district A. Legislators 4, 6, and 11 serve as counterfactuals for each other, as do
legislators 14, 15, 18, and 23.
The relative rank metric gives the percentile ranking of each legislator’s lottery number
relative to the legislators in their year-chamber-cohort-caucus district on a 0 to 1 scale.
Legislators assigned to 1 are the most senior in their year-chamber-cohort-caucus district group
(and thus can select the best committee assignment available to those in their caucus district
elected at the same time) and legislators with a 0 are the least senior. Likewise, a relative rank
value of 0.5 would mean that the legislator is at the 50th percentile and chooses in the middle of
her cohort.
We use relative rank as our main independent variable for the analysis because it has a
comparable meaning across the different year-chamber-cohort-caucus district groups in our
dataset. It is important to use a measure that is comparable across these groups because we are
not looking at one experiment in the analysis; instead, we are pooling and analyzing the results
9
from a series of many smaller experiments, one in each year-chamber-cohort-caucus district
group. As Table 1 highlights, the randomization only affects a legislators’ outcomes relative to
the outcomes of the members in their year-chamber-cohort-caucus district group. Having a
comparable measure facilitates this analysis. This is also the reason we drop legislators who have
no peers in their cohort and caucus district (such as legislator 13 in Table 1) from the analysis, as
these legislators have no counterfactual observations to which to compare.
We further account for the fact that we are pooling across the results of the ‘experiments’
that occur within these year-chamber-cohort-caucus district groups by including fixed effects for
year-chamber-cohort-caucus district groups in the analysis. This is important because we are
using data that covers a long time period. During the time period our data covers, Arkansas
implemented term limits and went from a solidly Democratic state to one where Republicans
became competitive with Democrats for control of the legislature. The fixed effects controls for
aspects of these temporally changes.
Contrasting Relative Rank And Typical Measures of Legislative Committee Assignments
The Arkansas’ randomized relative rank metric offers several distinct advantages over
the typical measures employed in studies of committee assignments by party leaders. To
appreciate the analogues between our measures and those employed in most studies, consider the
models of both processes depicted Box 1.
Box 1. Comparing Relative Rank and Typical Measures of Legislative Committee Quality
Typical Assignment Process
è
è
Party Leader
Desirability of
Outcomes
Regard
Committee for
Each Legislator
Endogenous,
Endogenous,
Measured
Measured with
Measured with
Proxy
Proxy
10
Randomized
Lottery Number
Exogenous,
Measurable
è
Arkansas Assignment Process
Desirability of
Committee for
Each Legislator
Unmeasured
è
Outcomes
Measured
The top half of Box 1 depicts the committee assignment process in most legislatures and
how scholars typically measure it. First, party leaders are thought to hold some legislators in
higher regard than others as a result of their service to the party, especially in the form of loyalty
on important roll call votes. This regard cannot be measured directly and is sometimes proxied
with party unity scores (e.g., Cox and McCubbins 2005). This party leader regard is also
endogenous to other aspects of the legislators’ career that might influence their assignments and
their legislative behavior, such as the safeness of their seat. Next, this regard is thought to
influence the quality of legislators’ committee assignments. Assignment quality cannot be
measured directly because some assignments may have greater value for some legislators than
others – for example, an assignment to the Agriculture committee may have significantly more
value to a legislator from a rural area than one from a city center. As discussed, assignment
quality is also potentially endogenous to other factors party leaders use to make choices
unrelated to their regard for the legislator – for example, legislators in marginal districts might be
more likely to request some committees than others. To study the committee assignment process,
existing studies typically examine the associations between various outcomes (e.g., re-election
margins) and these proxies for party leader regard and committee quality.
By contrast, consider the committee assignment process in Arkansas and the measures
available there. Instead of party leaders choosing which legislators to reward with their preferred
assignments, the randomized lottery determines which legislators get to select their preferred
11
assignments. Importantly, unlike the regard party leaders have for legislators, this first stage of
the assignment process can be directly measured in Arkansas (relative rank) and it is exogenous.
As in the typical setup, we cannot directly measure the desirability of each committee to
each legislator. However, as with other studies, we are not ultimately interested in the effects of
legislators winning particular committee assignments; we are interested in the effect of an in
increase in legislators’ ability to select their preferred assignment. In typical legislatures, one
would ideally estimate the effect of an exogenous increase in party leader regard for a member
on their outcomes; in Arkansas, there is an exogenous and measureable process which
determines whether legislators can choose from a broader and more desirable slate of
committees, relative rank. Because relative rank is randomly determined, we thus can measure
the impact of legislators having a better chance of getting their preferred assignments (the first
stage of the process) directly on their outcomes, similar to the exogenous dose of party leader
regard we would ideally deliver in other legislatures. Although there are undoubtedly other
factors that influence whether legislators achieve their goals or which committees they choose,5
these factors will be uncorrelated with the assignment mechanism in our data – the randomized
lottery – whereas in traditional data they may be highly correlated with the assignment
mechanism – the strategic decisions party leaders make. Likewise, as with party leader regard,
relative rank does not fully determine which committees legislators sit on – for example, junior
legislators with high relative ranks and who party leaders hold in high regard both cannot
displace senior legislators from choosing to return to the committees they have already served
on.
5
E.g., legislators both in Arkansas and elsewhere may decide to stay on less ‘objectively’ desirable committees to
develop seniority, etc.
12
Relative Rank Substantially Predicts Committee Assignments.
Although our design does not rely upon measures of each committee’s desirability to
each legislator, we first verify that committees in Arkansas are at least meaningfully
heterogeneous in their desirability to legislators by examining the mean relative rank of
legislators on different legislative committees. Panel (a) of Figure 1 shows the average relative
rank on the y-axis of the committee members in each committee in the Arkansas House (The
data comes from the Arkansas Legislative Digest and describes all sessions held between 1977
and 2011).
[INSERT FIGURE 1 HERE]
One of the advantages of studying committees in Arkansas is that we do not have to rely
on transfer requests to measure the value of committees (Stewart and Groseclose 1999) but can
directly look at legislators’ revealed preferences. And legislators’ revealed preferences
demonstrate that senior members (who have greater latitude over their committee assignments)
systematically prefer seats on some committees over others. Figure 1 displays this pattern. The
Insurance and Commerce Committee, for example, appears to fill up with the most senior
members at the beginning of the committee selection process. On the other hand, few members,
it seems, would serve on the Public Transport or Aging Committees if they had the choice to
serve on others. Panel (b) of Figure 1 shows the same pattern of results for the 35-member
Senate.
The same pattern holds within year-chamber-cohort-caucus district groupings. The scatter
plot in Figure 2 shows the relationship between the probability that legislators within each yearchamber-cohort-caucus district groupings served on one of the three most desirable committees,
given on the Y-axis, and their seniority number in their year-chamber-cohort-caucus district
group, given on the X-axis. The subgraphs in Figure 2 correspond to the size of the year-
13
chamber-cohort-caucus district group that each legislator serves in. The lines show the linear
best-fit for the data. As expected, the downward sloping lines in Figure 2 indicate that legislators
who are more senior within their year-chamber-cohort-caucus district are more likely to serve on
the more desirable committees (i.e. the committees preferred by the longest-serving members).
Moreover, the same top three committees consistently capture legislators’ interest. The opposite
pattern is observed for the least desirable committee (see Figure A1 in the Appendix).
Legislators randomized to higher (i.e., worse) seniority numbers within their cohort and caucus
district end up on committees disproportionately filled with others who have been similarly
(randomly) disadvantaged. If all committees in the Arkansas legislature were equally desirable, it
is very unlikely that we would observe this degree of systematic sorting.
[INSERT FIGURE 2 HERE]
Members who have higher relative ranks tend to systematically end up on top
committees as well. Although this pattern understates the importance of the randomization
because the value of each committee to each legislator cannot be directly measured, it is
encouraging for the validity of the design that legislators with better lottery numbers do
systematically end up on certain ‘top’ legislative committees much more often.
Figure 3 presents the lowess regression lines for legislators’ relative rank and their
likelihood of being on either a top committee in their chamber (panel a) or the least desirable
committee in the chamber (panel b), as determined by the raw seniority scores. As these graphs
show, Relative Rank substantially predicts the quality of legislators’ committee assignments.
Fully sixty percent of the legislators who are first in their cohort serve on one of the top
committees in the legislature, whereas those lower in the seniority rankings are about 15
percentage points less likely to serve on one of these committees. (Recall that legislators with
last pick by this metric still serve on top committees because the least senior members of the
14
most senior cohort still pick before any members of the next most senior cohort.) Panel (b)
shows that the effect for the least desirable committees is also strong. Table A2 in the
Supplementary Materials shows that these patterns are also highly statistically significant. The
Appendix also presents a randomization check verifying that the relative rank is independent
from observable characteristics of legislators’ districts.
[INSERT FIGURE 3 HERE]
Data and Dependent Variables
Our analysis uses 2,173 legislator-term observations from the period 1977-2011.6 We
analyze sixteen dependent variables related to four aspects of legislators’ careers and goals:
legislators’ electoral success, chamber leadership, policy productivity, and roll call voting.
For electoral goals, we used data from Carsey et al. (2007) and the Arkansas Secretary of
State’s website on whether each legislator won re-election, lost their primary re-election, lost
their general re-election, ran for or won higher office, retired, was opposed in the general
election, and was opposed in the primary election, as well as their general and primary election
vote shares. We collected the amount of campaign money that each incumbent raised from
www.followthemoney.org.
For chamber leadership, we collected data from the Arkansas Legislative Digest on
whether legislators served in party or official chamber leadership.
Policy productivity variables were only available for the years 2005-2008. For those
years we collected data from the Arkansas Legislative Digest on the number of bills legislators
filed and the number of bills they passed, a metric many other scholars have used to measure
6
There are 2,431 legislator-term observations during this period. However, only 2,173 of these observations are
used because some legislators were the only ones elected in their caucus district in their cohort, and thus were not
subject to any randomization, and because some committee assignment data was missing from 1977.
15
policy productivity and effectiveness.
Last, we used members’ roll call votes from 1997-2010 to construct three final dependent
variables about their voting and party loyalty: members’ extremity (based on their WNOMINATE score); the percentage of the time they vote with their party on roll calls where the
majority of Democrats opposed the majority of Republicans (Party Unity); and the percentage of
the time they vote with their party on roll calls where the majority of Democrats opposed the
majority of Republicans and their party lost the vote, i.e., when the majority of their party is
rolled (Party Unity (Losing Votes)).
Are Committee Assignments Valuable Enough To Meaningfully Incentivize Legislators?
It is clear that legislators prefer seats on some committees over others, but are these more
desirable committee seats so valuable that legislators would change their votes in order to win
them? Table 2 presents our estimates of the benefits legislators gain by obtaining their preferred
committee assignments. In all the regressions in the table the independent variable is Relative
Rank, the scaled randomized seniority of legislators within their year-chamber-cohort-caucus
district that allows them to pick from a much larger and more desirable set of committees.
Because Relative Rank varies from 0 to 1, the coefficient on this variable indicates the estimated
difference between the most and least senior member within each cohort – that is, between the
cohort members who have the most and least choice in their assignments. Recall as described in
the previous section that these limiting cases are roughly analogous to the situations in which a
hypothetical party leader wished to reward a loyal legislator with the best available committee
assignment that more senior members had not already taken or, alternatively, consign a disloyal
16
one to the last remaining assignment after all others legislators’ wishes were granted.7
The dependent variables, each described in the previous section, are listed under each of
the headings in Table 2. For each outcome we present the results from a regression without any
fixed effects and the results from a regression with fixed effects for year-chamber-cohort-caucus
district (i.e., the groups within which the randomizations occur). The last two rows under each
dependent variable in Table 2 describe the number of observations and, when applicable, the
number of fixed effects. As discussed, these dependent variables correspond to the key goals that
scholars near-unanimously posit desirable committee assignments help legislators achieve.
Table 2 shows that legislators’ relative rank does not have a statistically significant effect
on any measures of their election outcomes. Legislators who have their pick of desirable
committee assignments are not meaningfully more likely to win their primary or general election
bids, raise campaign money, run for or win for higher office, deter opponents, or increase their
vote share.8 They are also not noticeably more likely to be successful in achieving progressive
ambition goals. When we spoke with Arkansas legislators, they told us that because of termlimits in the state, progressive ambition is one of their most salient goals (see also Kousser
2005). However, as Table 2 shows, we fail to reject the null hypothesis that legislators who
choose their committee assignments first are just as likely to either run for or win higher office.
The variable we examine related to chamber leadership – whether the legislator is elected
to serve as a party leader in their chamber or a presiding office in their chamber – is also
insignificant. We also find that legislators are no more likely to write nor pass bills as a result of
their seniority.
7
Legislators can be expected to choose their most preferred available choice since they choose for themselves (i.e.,
in a “serial dictatorship” arrangement; see Satterthwaite and Sonnenschein 1981).
8
This finding is consistent with qualitative evidence from Hall (1996), who finds that lawmakers generally do not
see committee participation as an especially electorally rewarding activity.
17
Last, we find no appreciable effect of assignments on whether legislators vote with their
parties or vote more extremely. One interpretation of this effect is that legislators with better
committee assignments are no more likely to influence the agenda such that they support it;
another is that legislators do not vote differently simply by virtue of their assignments.9
In their totality, the results of the analysis are clear and highly surprising in light of
decades of scholars’ conventional wisdom about the value of desirable committee assignments.
Across all thirty-four estimates for the seventeen dependent variables, we find no statistically
significant effects for a legislators’ seniority on the outcomes of interest we identified (with a
generous threshold of p < 0.10). Further, our estimates are based on a large number of
observations and have substantively small standard errors. For example, the 95% confidence
interval for the estimate of the decreased probability that a legislator loses a general election
because of their seniority extends only to 1.2 percentage points. In summary, the benefits
legislators may reap from having an upper hand in the committee assignment process appear to
be at best meager.
We also tested whether the results varied by legislators’ tenure in office. It is possible that
we are missing the benefits committees hold because the benefits are isolated to specific times in
their career and the current analysis misses that because it averages over their entire career. To
examine this hypothesis, we tested whether the effect of relative rank varies with tenure by
rerunning the models and including dummy variable for legislators in the second term and
another for legislators serving in their third-plus term (i.e., their third, fourth, fifth, etc. term) and
9
This finding also rules out an alternative explanation for the null findings, namely that legislators do receive
electoral benefits from more attractive committee assignments but use this additional ‘political capital’ to vote more
extremely. In this way, committee membership might grant members additional leeway to vote against their
constituents’ preferences (Cain, Ferejohn, Fiorina 1987, p. 87). However, the results show that legislators with more
attractive committee assignments are neither more likely to be extreme nor more likely to vote with their party.
Likewise, the consistent null results across the large number of variables we tested suggests that legislators do not
use ‘political capital’ from their committee assignments in order to help them achieve other goals.
18
interaction terms between these dummy variables and relative rank. The results of the analysis
are presented in Table A3 of the Supplementary Materials. In a few cases the effect of relative
rank is significant, but in those few cases it goes in the wrong direction. These few significant
negative effects are likely due to chance, as would be expected when testing a large number of
hypotheses.
These results speak to longstanding debates about whether party leaders can use the
committee assignment process to meaningfully incentivize party loyalty (see review in previous
section). Although legislators clearly do prefer some assignments to others and thus must see
some degree of preferences over assignments, these benefits appear to be at most quite small. On
the other hand, legislators face quite significant electoral costs for voting out of step with their
constituents (e.g., Fiorina 1974; Canes-Wrone, Brady, and Cogan 2002; Masket and Greene
2011). Given the substantial electoral costs legislators are thought to face for voting out of step,
our evidence thus suggests that the committee assignment process is unlikely to plausibly form
an important basis for the formidable power over legislators’ roll call votes legislative parties are
thought to exercise.
External Validity
One may well wonder whether results from one state’s legislature could meaningfully
speak to general theories of legislative organization. Are committees and parties in the Arkansas
legislature similar enough to those in other legislatures that scholars should take interest in these
results, or is Arkansas so different that the results shed little generalizable light? We argue for
the external validity of our results by briefly discussing how the committee and party system in
Arkansas shares the key institutional details central to the theories of committees with which we
engage.
19
First, the majority party in Arkansas seems to exercise levels of power that are
comparable to levels of power enjoyed in other state legislatures. In particular, Anzia and
Jackman (2012), calculate the average majority party roll rate – a key barometer of party power
(e.g., Cox and McCubbins 2005) – in all 99 state legislative chambers in the United States and
find that both of Arkansas’ chambers are typical of other state legislatures, occupying positions
solidly in the middle of the pack (see Figure A2).
More importantly, the features that form the basis for committee power in legislative
theories also exist in Arkansas. Committees have sole jurisdiction over large policy areas,
supervise regulatory agencies, are responsible for doling out substantial sums of state money in
their areas of jurisdiction, and are de facto veto points for legislation in these areas. Discharge
petitions circumventing committee veto power are exceedingly rare in Arkansas, with legislators
and staff we spoke with only able to recall two occasions when a discharge petition has been
successful in the last two decades.
Legislators in Arkansas we interviewed also described procedures by which legislators
can kill legislation they do not support in their committees – for example, by leaving the room
during the vote (because nays and absences count identically). When a committee fails to pass a
bill, legislators in Arkansas consider it “dead.”
Likewise, committees in Arkansas undertake significant regulatory oversight and help
appropriate the state’s $4.4 billion annual budget, which include a number of opportunities for
members to steer money towards their districts.
Arkansan legislators should also fear electoral rewards and punishments to the same
degree their colleagues elsewhere do. Like members of other legislatures they are much more
likely to win than to lose, although more than 20% of legislator-years in our data feature
legislators losing or retiring. And, as with other legislatures, simply because many legislators do
20
win re-election does not mean they do not fear losing or that their seats are automatically safe
(e.g., Erikson 1976).
As with any data in the social sciences, results from this one setting are of course not
ultimately dispositive about other contexts. For example, during the period covered by our data,
Arkansas implemented legislative term limits. It may be that these results do not speak to
contexts where there are no term limits (e.g., Congress). It is also the case that the state went
from being a solidly Democratic state to one where Republicans are now competitive with
Democrats in battles over legislative control. These differences may have decreased the value of
committee assignments in Arkansas. Although we cannot definitely speak to what would have
occurred had Arkansas not implemented term limits when it did, we reran the main analysis (i.e.,
Table 2) with interaction terms for each decade to see if in one of the early decades relative rank
had a significant effect. The results, which are provided in the supplementary materials (see
Table A4), show that the effects do not change much at all. The variable relative rank fails to
achieve statistical significance as a predictor for thirteen of the fourteen beneficial outcomes we
measure. Further, the coefficient on relative rank actually points in the wrong direction for the
one outcome for which it does achieve statistical significance (vote share in the general election).
Having a better committee choice set does not appear to benefit legislators and this pattern is
consistent across the previous three decades.
Discussion
Scholars’ leading explanation for high levels of party unity in American legislatures has
long been that parties exploit their control over access to seats on powerful committees to
incentivize rank-and-file members to be “more responsive to both the party’s leadership and
goals” (Cox and McCubbins 1993, p. 182): desirable committee assignments are typically held to
21
present considerable benefits for legislators, and legislative leaders are thus believed to use the
excludable nature of these benefits as carrots (and sticks) to achieve party unity (e.g., Shepsle
1978, p. 35; Smith 2000, p. 62; Stratmann 2000, p. 666; Snyder and Groseclose 2000, p. 194; see
section 1 for review). Unfortunately, existing studies have been unable to distinguish an
inducement view of the committee assignment process from observationally equivalent
alternatives, a challenge we helped address with data from the Arkansas legislature, where party
leaders play no role in the committee assignment process and where a randomized lottery
allowed us to estimate the causal effect of legislators’ membership on their desired committees.
The results showed that Arkansas state legislators are not significantly more likely to
succeed in their re-election campaigns, raise money, influence policy, become leaders in their
chambers, or vote differently by virtue of having a better selection of committee assignments.
Legislators likely do reap some benefits for winning their preferred assignments, but our
estimates’ confidence intervals ruled out the existence of all but substantively small such
benefits. These results suggest that the benefits of winning preferred committee assignments are
at most too small to plausibly form the basis of party power, especially when considered relative
to the significant electoral costs legislators face for voting out of step with their constituents
(e.g., Fiorina 1974; Canes-Wrone, Brady, and Cogan 2002; Masket and Greene 2011). To be
clear, our data do not support the conclusion that better committee assignments have no value to
legislators – legislators clearly have some degree of preference over committees, suggesting they
perceive at least a small advantage in sitting on some committees over others (see Figures 1 and
2). However, the benefits from these committees seem at most far too limited to form a
significant foundation for a workable theory of party power.
Do Party Leaders Use Control Over Assignments to Control Legislators or Committees?
22
Given that desirable committee assignments do not appear to be lucrative enough to
induce legislators to be loyal to their parties, why do party loyalists tend to fill the most powerful
legislative committees in most legislatures (e.g., Cox and McCubbins 1993; Kanthak 2004;
Hedlund et al. 2011)? One reason may be that party leaders use their control over the committee
assignment process in most legislators to stack powerful committees with loyalists because they
want dependable partisans to control these key legislative institutions. Arkansas provides a
unique setting to examine whether party leaders are responsible for the presence of loyalists on
powerful committees in most legislatures because they do not control assignments there,
allowing us to shed some light on whether loyal legislators naturally select on to these
committees or if the overt intervention of leaders is necessary to achieve these patterns of
committee membership.
To evaluate what kinds of committees the most loyal legislators tend to in the absence of
party leader control, we regress the same dependent variables dealing with how legislators vote
on roll calls as shown in Table 2 on a dummy variable for whether the legislator sits on one of
these top three committees. For the sake of comparability, we limit the sample to legislators on
whom we also had seniority data (though the results still hold when including all legislators).
These results are shown in Table 3. For consistency with Table 2, we estimate specifications
with and without the fixed effects for year-chamber-cohort-caucus district groups.
[INSERT TABLE 3 HERE]
A purely self-selection based account of the presence of party loyalists on powerful
committees in other legislators would predict that loyal legislators would tend to fill the most
desirable committees in Arkansas as well. However, most of the specifications in Table 3 are
statistically significant in the opposite direction, indicating that disloyal legislators tend to fill the
most attractive committees in this setting where party leaders play no role in the assignment
23
process. Such patterns of self-selection go in the exact opposite direction of what scholars
reliably find in other legislatures (e.g., Cox and McCubbins 1993; Kanthak 2009).
In summary, our results lend support to the argument that party leaders are responsible
for the presence of their loyalists on powerful committees, but that they do so to control
committees, not to control legislators. Although our data is limited to one legislature, it suggests
that party leaders’ well-documented attention to their members’ loyalty during the committee
assignment process10 may be best explained by their desire to control committees by stacking
them with loyal legislators, not a quest to generate incentives for legislators to remain loyal. For
example, as was starkly illustrated with House Speaker John Boehner’s 2011 placement of
staunch conservative House members Jeb Hensarling, Dave Camp, and Fred Upton on the socalled ‘budget supercommittee’ that was charged with reducing the federal debt, legislative
parties recognize that loyal partisans on key committees play a key role in maintaining control
over the agenda and shaping legislation – but it is clear that Boehner primarily placed
Hensarling, Camp, and Upton on this key committee in order to influence what the committee
did, not to meaningfully incentivize other members to vote with the party.
How parties succeed in usurping the power of their legislatures to pass their preferred
policies and control the agenda is an (increasingly) important question. Scholars have developed
a number of compelling accounts for this pattern, including, among others, majority party control
over the legislative agenda (e.g., Anzia and Jackman 2012), primary competition (e.g., Burden
2004), heterogeneity of constituent preferences (Harden and Carsey 2012), willful cooperation of
extreme legislators themselves (e.g., van Houweling 2013), and the role of extra-legislative
interest group organizations (Masket 2007). Our findings suggest that scholars may want to look
10
For example, Manley (1970), Hinckley (1983), and Sinclair (1995) all document in qualitative interviews and
from participant-observation that party leaders diligently consider legislators’ record of party loyalty when making
assignments.
24
beyond party control over the committee assignment process to other explanations like these to
understand why legislators so frequently spurn their median constituents to vote with their
legislative copartisans.11
References
Adler, E. Scott, and John S. Lapinski. 1997. “Demand-Side Theory and Congressional
Committee Composition: A Constituency Characteristics Approach.” American Journal
of Political Science 41 (3): 895-918.
Anzia, Sarah F. and Molly C. Jackman. 2012. “Legislative Organization and the Second Face of
Power: Evidence from US State Legislatures.” Forthcoming, Journal of Politics.
Bullock, Charles S. III. 1972. “Freshman Committee Assignments and Re-election in the United
States House of Representatives.” American Political Science Review 66(3): 996-1007.
Bullock, Charles S. III. 1976. “Motivations for Congressional Committee Preferences: Freshmen
of the 92nd Congress.” Legislative Studies Quarterly 1: 201-212.
Burden, Barry C. 2004. “Candidate Positioning in US Congressional Elections.” British Journal
of Political Science 34: 211-227.
Cain, Bruce E., John Ferejohn, and Morris P. Fiorina. 1987. The Personal Vote: Constituency
Service and Electoral Independence. Cambridge: Harvard University Press.
Canes-Wrone, Brandice, David W. Brady, and John F. Cogan. 2002. “Out of Step, Out of Office:
Electoral Accountability and House Members’ Voting.” American Political Science
Review 96 (1): 127-140.
Carsey, Thomas M., and Barry Rundquist. 1999. “Party and Committee in Distributive Politics:
11
Likewise, rather than focusing on the benefits that come from committee assignments, the results suggest that
researchers might do better to focus on the benefits legislators and parties derive from the committee system (e.g.,
Krehbiel 1991; Gilligan and Krehbiel 1997).
25
Evidence from Defense Spending.” Journal of Politics 61 (4): 1156-1169.
Carsey, Thomas M., William D. Berry, Richard G. Niemi, Lynda W. Powell, and James M.
Snyder. 2007. STATE LEGISLATIVE ELECTION RETURNS, 1967-2003.
Crain, W. Mark, and John T. Sullivan. 1997. “Committee Characteristics and Re-Election
Margins: An Empirical Investigation of the US House.” Public Choice 93 (3): 271-285.
Crook, Sara Brandes, and John R. Hibbing. 1985. “Congressional Reform and Party Discipline:
The Effects of Changes in the Seniority System on Party Loyalty in the US House of
Representatives.” British Journal of Political Science 15 (2): 207-226.
Coker, David C. and W. Mark Crain. 1994. “Legislative Committees as Loyalty-Generating
Institutions.” Public Choice 81: 195-221.Cox, Gary W. and Mathew D. McCubbins.
1993. Legislative Leviathan: Party Government in the House. University of California
Press.
Cox, Gary W. and Mathew D. McCubbins. 2005. Setting the Agenda: Responsible Party
Government in the US House of Representatives. Cambridge.
Dow, Jay K., James W. Endersby, and Charles E. Menifield. 1998. “The Industrial Structure of
the California Assembly: Committee Assignments, Economic Interests, and Campaign
Contributions.” Public Choice 94 (2): 67-83.
Erikson, Robert S. 1976. “Is There Such a Thing as a Safe Seat?” Polity 26(4): 531-546.
Fenno, Richard F. 1973. Congressmen in Committees. Little Brown.
Fiorina, Morris. 1974. Representatives, Roll Calls, and Constituencies. Lexington, MA: D.C.
Heath.
Fowler, Linda L., Scott R. Douglass, and Wesley D. Clark, Jr. 1980. “The Electoral Effects of
House Committee Assignments.” Journal of Politics 42 (1): 307-319.
Fowler, Anthony, and Andrew B. Hall. 2013. “Long Term Consequences of Elections.” Working
26
Paper, Harvard University.
Gaines, Brian J., Timothy P. Nokken, and Collin Groebe. 2012. “Is Four Twice as Nice as Two?
A Natural Experiment on the Electoral Effects of Legislative Term Length.” State
Politics & Policy Quarterly 12 (1): 43-57.
Gertzog, Irwin N. 1976. “The Routinization of Committee Assignments in the US House of
Representatives.” American Journal of Political Science 20: 693-712.
Gilligan, Thomas W., and Keith Krehbiel. 1997. “Specialization Decisions Within Committee.”
Journal of Law, Economics, and Organization 13 (2): 366-386.
Grier, Kevin B. and Michael C. Munger. 1993. “Comparing Interest Group PAC Contributions to
House and Senate Incumbents, 1980-1986.” Journal of Politics 55(3): 615-643.
Grier, Kevin B. and Michael C. Munger. 1991. “Committee Assignments, Constituent
Preferences, Campaign Contributions.” Economic Inquiry 29 (1): 24-43.
Grimmer, Justin. 2012. “Appropriators Not Position Takers: The Distorting Effects of Electoral
Incentives on Congressional Representation.” Forthcoming, American Journal of
Political Science.
Grimmer, Justin, and Eleanor Powell. 2013. “Congressmen in Exile: The Politics and
Consequences of Involuntary Committee Removal.” Forthcoming, Journal of Politics.
Hall, Richard L. 1996. Participation in Congress. Yale University Press.
Harden, Jeffrey J., and Thomas M. Carsey. 2012. “Balancing Constituency Representation and
Party Responsiveness in the US Senate: The Conditioning Effect of State Ideological
Heterogeneity.” Public Choice 150 (1-2): 137-154.
Heberlig, Eric. S. 2003. “Congressional Parties, Fundraising, and Committee Ambition.”
Political Research Quarterly 56: 151-162.
Hedlund, Ronald D. and Samuel C. Patterson. 1992. “The Electoral Antecedents of State
27
Legislative Committee Assignments.” Legislative Studies Quarterly 17 (4): 539-559.
Hedlund, Ronald D., Rob A. DeLeo, and David P. Hedlund. 2011. “Reading the Tea Leaves:
Legislative Risk-Taking During the Committee Request and Assignment Process.” Paper
presented at the 2011 State Politics and Policy Conference, Dartmouth University.
Hedlund, Ronald D., Kevin Cooms, Nancy Martorano, and Keith E. Hamm. 2011. “Partisan
Stacking on Legislative Committees.” Legislative Studies Quarterly 34(2): 175-191.
Kanthak, Kristin. 2004. “Exclusive committee assignments and party pressure in the US House
of Representatives.” Public Choice 121: 391-412.
Kanthak, Kristin. 2009. “US State Legislative Committee Assignments and Encouragement of
Party Loyalty: An Exploratory Analysis.” State Politics and Policy Quarterly 9(3): 284303.
Katz, Jonathan N., and Brian R. Sala. 1996. “Careerism, Committee Assignments, and the
Electoral Connection.” American Political Science Review 90: 21-33.
Kellerman, Michael, and Kenneth A. Shepsle. 2009. “Congressional Careers, Committee
Assignments, and Seniority Randomization in the US House of Representatives.”
Quarterly Journal of Political Science 4: 87-101.
Kousser, Thad. 2005. Term Limits and the Dismantling of State Legislative Professionalism.
Cambridge.
Krehbiel, Keith. 1991. Information and Legislative Organization. University of Michigan Press.
Krehbiel, Keith. 1993. “Where’s the Party?” British Journal of Political Science 23 (2): 235-266.
Krehbiel, Keith. 2003. “The Coefficient of Party Influence.” Political Analysis 2003 11(1): 95103.
Leighton, Wayne A. and Edward J. Lopez. 2002. “Committee Assignments and the Cost of Party
Loyalty.” Political Research Quarterly 55 (1): 59-90.
28
Maltzman, Forrest. 1997. Competing Principals: Committees, Parties, and the Organization of
Congress. Ann Arbor: University of Michigan Press.
Manley, John F. 1970. The Politics of Finance: The House Committee on Ways and Means.
Little Brown.
Masket, Seth E. 2007. “It Takes an Outsider: Extralegislative Organization and Partisanship in
the California Assembly, 1849-2006.” American Journal of Political Science 51(3): 482497.
Masket, Seth E. and Steven Greene. 2011. “When One Vote Matters: The Electoral Impact of
Roll Call Votes in the 2010 Congressional Elections.” Paper Presented at the 2011
Meeting of the Midwest Political Science Association.
Masters, Nicholas A. 1961. “Committee Assignments in the House of Representatives.”
American Political Science Review 55 (2): 345-357.
Mayhew, David R. 1974. Congress: The Electoral Connection. Yale University Press.
McCarty, Nolan, Keith T. Poole, and Howard Rosenthal. 2001. “The Hunt for Party Discipline in
Congress.” American Political Science Review 95 (3): 673-687.
McCarty, Nolan, Keith T. Poole, and Howard Rosenthal. 2006. Polarized America: The Dance
of Ideology and Unequal Riches. MIT Press.
Milyo, Jeffrey. 1997. “Electoral and Financial Effects of Changes in Committee Power: The
Gramm-Rudman-Hollings Budget Reform, the Tax Reform Act of 1986, and the Money
Committees in the House.” Journal of Law and Economics 40 (1): 93-111.
Roberts, Jason M. and Steven S. Smith. 2003. “Procedural Contexts, Party Strategy, and
Conditional Party Voting in the US House of Representatives, 1971-2000.” American
Journal of Political Science 47(2): 305-317.
Rohde, David W. and Kenneth A. Shepsle. 1973. “Democratic Committee Assignments in the
29
House of Representatives: Strategic Aspects of a Social Choice Process.” American
Political Science Review 67: 889-905.
Romer, Thomas, and James M. Snyder. 1994. “An Empirical Investigation of the Dynamics of
PAC Contributions.” American Journal of Political Science 38 (3): 745-769.
Satterthwaite, Mark and Hugo Sonnenschein. 1981. “Strategy-proof Allocation Mechanisms at
Differentiable Points.” Review of Economic Studies 48(4): 587-597.
Shepsle, Kenneth A. 1978. The Giant Jigsaw Puzzle: Democratic Committee Assignments in the
Modern House. University of Chicago Press.
Sinclair, Barbara. 1995. Legislators, Leaders, and Lawmaking. Baltimore: Johns Hopkins
University Press.
Smith, Steven S., and Christopher J. Deering. 1983. “Changing Motives for the Committee
Preferences of New Members of the United-States House.” Legislative Studies Quarterly
8: 271-281.
Smith, Steven S. and Christopher J. Deering. 1984. Committees in Congress. Washington, DC:
Congressional Quarterly Press.
Smith, Steven S. 2000. Party Influence in Congress. Cambridge.
Snyder, James M., Jr., and Tim Groseclose. 2000. “Estimating Party Influence in Congressional
Roll-Call Voting.” American Journal of Political Science 44(2): 193-211.
Stewart, Charles III, and Tim Groseclose. 1999. “The Value of Committee Seats in the United
States Senate, 1947-91.” American Journal of Political Science 43(3): 963-973.
Stratmann, Thomas. 2000. “Congressional Voting Over Legislative Careers: Shifting Perceptions
and Changing Constraints.” American Political Science Review 94 (3): 665-676.
Weingast, Barry R. and William J. Marshall. 1988. “The Industrial Organization of Congress; or,
Why Legislatures, Like Firms, Are Not Organized as Markets.” Journal of Political
30
Economy 96(1): 132-163.
Westefield, Louis P. 1974. “Majority Party Leadership and the Committee System in the House
of Representatives.” American Political Science Review 68 (4): 1593-1604.
31
Table 1. Hypothetical Example of how Relative Rank is computed
(a)
(b)
Grouped by Cohort and Seniority
Grouped by Cohort Caucus District, and Seniority
Term
Random
Caucus
Term
Random
Caucus
Relative
Number Seniority
District
Number
Seniority District
Rank
(Cohort)
(Cohort)
2
1
D
2
4
A
1
2
2
B
2
6
A
.5
2
3
B
2
11
A
0
2
4
A
2
2
B
1
2
5
B
2
3
B
.67
2
6
A
2
5
B
.33
2
7
C
2
9
B
0
2
8
D
2
7
C
1
2
9
B
2
10
C
0
2
10
C
2
1
D
1
2
11
A
2
8
D
0
1
12
D
1
14
A
1
1
13
B
1
15
A
.67
1
14
A
1
18
A
.33
1
15
A
1
23
A
0
1
16
C
1
13
B
1
17
C
1
16
C
1
1
18
A
1
17
C
.67
1
19
D
1
20
C
.33
1
20
C
1
25
C
0
1
21
D
1
12
D
1
1
22
D
1
19
D
.75
1
23
A
1
21
D
.5
1
24
D
1
22
D
.25
1
25
C
1
24
D
0
Notes: This table illustrates how relative rank is calculated using a hypothetical 25-member
Arkansas House populated with legislators who were either just elected or are serving their
second term.
32
Table 2. Effect of Seniority within Cohort (Relative Rank) on Outcomes of Interest (OLS)
Electoral Goals
Win Reelection
Lose Primary
Lose General
Run for Higher Office
F.E.
No F.E.
F.E.
No F.E.
F.E.
No F.E.
F.E.
No F.E.
Coeff.
0.012
0.026
-0.001
-0.004
0.001
0.001
-0.002
-0.002
Std.
(0.021)
(0.029)
(0.010)
(0.010)
(0.006)
(0.006)
(0.012)
(0.012)
Error
# Obs.
1,875
1,875
1,875
1,875
1,875
1,875
2,084
2,084
# F.E.
433
433
433
441
-
Coeff.
Std.
Error
# Obs.
# F.E.
Coeff.
Std.
Error
# Obs.
# F.E.
Opposed in General
F.E.
No F.E.
0.014
0.010
(0.019)
(0.020)
1,875
433
1,875
-
Chamber Goal
Serve as Chamber
Leader
F.E.
No F.E.
-0.002
-0.002
(0.013)
(0.013)
2,084
441
2,084
-
Opposed in Primary
F.E.
No F.E.
0.005
0.000
(0.020)
(0.021)
1,875
433
1,875
-
Vote Share in General
F.E.
No F.E.
-0.012
-0.016
(0.031)
(0.020)
207
133
207
-
Policy Productivity Goals
Number of Bills
Number of Bills Passed
Filed
F.E.
No F.E.
F.E.
No F.E.
-2.83
-2.54
-0.74
-0.56
(1.85)
(2.19)
(1.25)
(1.51)
264
30
264
-
264
30
264
-
Vote Share in Primary
F.E.
No F.E.
-0.026
-0.006
(0.038)
(0.022)
223
163
223
-
Win Higher Office
F.E.
No F.E.
0.008
0.008
(0.009) (0.009)
2,084
441
2,084
-
1,043
-
1,875
433
1,875
-
Money Raised
F.E.
No F.E.
-9164
-2895
(6512)
(6327)
453
61
453
-
Roll Call Voting
Party Unity
Party Unity
(Losing Votes)
F.E.
No F.E.
F.E.
No F.E.
0.002
0.001
-0.005
-0.003
(0.013)
(0.014) (0.022) (0.024)
1,043
133
Retire
F.E.
No F.E.
-0.009
-0.018
(0.018) (0.023)
1,001
133
1,001
Extremity
(W-Nominate)
F.E.
No F.E.
-0.005
-0.005
(0.032) (0.032)
1,040
133
1,040
-
Notes: Fixed effects refer to the groups in which the randomization takes place (i.e. for each year-chamber-cohort-caucus district
group). The independent variable for all regressions, relative rank, is the scaled random seniority rank of each legislator within their
randomization group. The variable ranges from 0 to 1, with legislators assigned to 1 as the most senior. Coefficients represent the
estimated effects of being the most senior member instead of the least senior member. No outcomes are significant at the 0.10 level. Ns
differ in regressions with dependent variables for which data is not available for all years.
33
Table 3. Association Between Committee Attractiveness and Party Loyalty (OLS)
Dependent Variable
Party Unity
Extremity
Party Unity
(Losing Votes)
(W-Nominate)
Fixed Effects Used?
Yes
No
Yes
No
Yes
No
Dummy for On a Top -0.027** -0.017^ -0.038*
-0.027^
-0.045*
-0.018
3 Committee
(0.009)
(0.009)
(0.016)
(0.016)
(0.024)
(0.023)
Constant
0.764**
0.563**
0.449**
0.436**
(0.006)
(0.011)
(0.016)
(0.016)
N
1,043
1,043
1,001
1,001
1,040
1,040
# of Fixed Effects
133
133
133
Notes: Fixed effects refer to the groups in which the randomization takes place (i.e. for each
year-chamber-cohort-caucus district group). ^Sig. at the 0.10 level (two-tailed), *Sig. at the 0.05
level (two-tailed), **Sig. at the 0.01 level (two-tailed).
34
Figure 1. Average Seniority Scores on House and Senate Committees
Notes: The graph gives the average seniority score (with a 95 percent confidence interval) of the
members serving on the standing committees in the Arkansas state House (in panel a) and Senate
(b) from the period 1977-2011. The seniority scores range from 1 to 100 for the House and from
1 to 35 for the Senate.
35
Figure 2. Probability of Being on a Top Committee, by Seniority Number (Sub-graphs by
cohort size)
Notes: Each sub-graph corresponds to the size of the year-chamber-cohort-caucus district group
that each legislator serves in. Each dot represents the percent of people with that lottery number
in their cohort that serves on one of the top committees in the chamber.
36
Figure 3. Probability of Being on Best and Worst Committees, by Relative Lottery Number
Notes: This presents the locally weighted regression (lowess) to estimate the predicted
probability that a legislator serves on one of the top committees (panel a) and the least desirable
committee (panel b) in the chamber based on their relative rank in their year-chamber-cohortcaucus district group.
37
Appendix for “How Do Committee Assignments Facilitate Legislative Party Power?
Evidence from a Randomized Experiment in the Arkansas State Legislature”
Randomization Check: Balance in Covariates Across Groups
We conducted a randomization using covariates of interest available from the US Census
were constant across our randomized treatment. Specifically, we tested whether legislators who
were assigned higher seniority were more likely to come from districts that had different median
ages, median household incomes, population percentage black, Asian, and Hispanic, and
population percent rural. We also checked the legislator’s partisanship.
Unfortunately, the US Census only began providing this legislative district level data
beginning with the 2000 Census. However, the seniority selection process has remained the same
throughout the past several decades so we do not expect that our results would have differed if
we had access to such data for previous decades.
Table A1. Balance of Covariates Across Relative Rank and Seniority Score
Independent Variables
Dependent Variable: Relative
Dependent Variable:
Rank
Absolute Seniority Rank
Legislator is a Democrat
0.023
-4.732
(0.031)
(2.845)
Median Age
-0.001
-0.018
(0.004)
(0.322)
Median Household Income (in
0.007
0.258
$10,000s)
(0.019)
(1.699)
Black Percent
-0.128
-4.161
(0.087)
(7.834)
Asian Percent
-1.281
-32.54
(1.485)
(133.4)
Hispanic Percent
-0.383
-6.980
(0.256)
(22.98)
Rural Percent
-0.022
-5.063
(0.065)
(5.811)
Constant
0.542**
48.63**
(0.175)
(15.78)
2
R
.0120
.0094
N
673
673
Notes: In the first column, the dependent variable is the scaled random seniority rank of each
legislator within their caucus district. The variable ranges from 0 to 1, with legislators assigned
to 1 as the most senior. In the second column, dependent variable is the seniority rank of each
legislator within their caucus district. The variable ranges from 1 to 100 in the House and 1 to 35
in the Senate, with legislators assigned to 1 as the most senior. No results are significant at the
0.10 level.
The two columns in Table A1 display the results and show that both the Relative Rank
metric and legislators’ seniority numbers themselves are independent of the characteristics of
their districts. F tests show that both relative rank (F(7,665) = 1.15, p = .32) and the raw seniority
scores F(7,665) = 0.90, p = .51) are unrelated to these characteristics. This gives us additional
38
confidence that the randomization was successful and no other confounding factors lead some
legislators to gain better committee assignments than others.
The Desirability of Legislators’ Committee Assignments and their Relative Rank
The regression results displayed in Table A2 show that the pattern in Figure 3 holds up
quantitatively as well, with the most senior members in a cohort about 15 percentage points more
likely to sit on one of the top committees and members with the lowest seniority about 13
percentage points more likely to serve on the least desirable committee. Both results are
statistically significant at the 0.01 level. The two columns for each set of regressions show that
these effects are robust to the inclusion of fixed effects for each chamber-cohort-caucus district
group.
We also test our assumptions by assigning each legislator a score that corresponds to the
average seniority number on that legislators’ most desirable committee. To do so, we first assign
each committee in each session a score corresponding to the average seniority number on that
committee. We then calculate the desirability of best committee, the average seniority on the
committee with the highest such average of all the committees on which a legislator sits. We
finally rescale this metric from 0 to 1, so that 1 corresponds to the most desirable committee in
any session and 0 corresponds to the least desirable committee in any session. Table A2 shows
that legislators’ Relative Rank again substantially predicts how desirable their best committee is.
Table A2. Effect of Relative Rank on Likelihood of Serving on Desirable Committees
(OLS)
Dependent Variable
Fixed Effects?
Relative Rank
Constant
N Observations
Number of Fixed
Effects
On Top Committee
Yes
No
0.15**
0.15**
(0.03)
(0.03)
0.41**
(0.02)
2,084
441
2,084
-
On Least Desirable
Committee
Yes
No
-0.14**
-0.14**
(0.02)
(0.02)
0.26**
(0.01)
2,084
441
2, 084
-
Desirability of Best
Committee
Yes
No
0.06**
0.06**
(0.01)
(0.01)
0.49**
(0.01)
2,054
438
2,054
-
Notes: The independent variable for all regressions, relative rank, is the scaled random seniority
rank of each legislator within their caucus district. The variable ranges from 0 to 1, with
legislators assigned to 1 as the most senior. Coefficients represent the estimated effects of being
the most senior member instead of the least senior member. Fixed effects are used for the groups
in which the randomization takes place (i.e. for each chamber-cohort-caucus district group).
^Sig. at the 0.10 level (two-tailed), *Sig. at the 0.05 level (two-tailed), **Sig. at the 0.01 level
(two-tailed).
These results validate our assumption that a legislators’ Relative Rank constitutes a large,
exogenous shock to the desirability of their assignments. However, note that these aggregate
measures only capture the aspects of committee desirability that are common to all legislators,
whereas many committees (e.g. Agriculture) are likely far more attractive to some legislators
than to others. It is therefore likely that these results significantly understate the degree to which
a legislator’s Relative Rank affects the attractiveness of her assignments.
39
Revealed Preference for the Least Desirable Committee
Figure A1 shows the relationship between the probability that legislators served on the
least desirable committees, given on the Y-axis, and their seniority number in their chambercohort-caucus district group, given on the X-axis. The subgraphs in Figure A1 correspond to the
size of the chamber-cohort-caucus district group that each legislator serves in. The lines show the
linear best-fit for the data. The upward sloping lines in Figure A1 indicate that legislators
randomly assigned less senior lottery numbers are more likely to serve on the least desirable
committee in their chamber.
Figure A1. Probability of Being on Least Desirable Committee, by Seniority Number (Subgraphs by cohort size)
Notes: Each sub-graph corresponds to the size of the chamber-cohort-caucus district group that
each legislator serves in. Each dot represents the percent of people with that lottery number in
their cohort that serves on the least desirable committee in the chamber. The lines show the
linear best fit for the data.
Heterogeneous Treatment Effects by Tenure and Decade
We conducted several additional analyses that examined whether incumbent tenure or
decade moderated the effect of relative rank. The results of the analysis are presented in Tables
A3 (tenure) and A4 (decade). For these analyses we estimated regressions that included dummy
40
variable for the characteristics of interest (either incumbent tenure or decade) and the interaction
between these variables and relative rank. The models did include an intercept term but for space
reasons we do not present it. We also present the results both with and without the fixed effects.
Fixed effects refer to the groups in which the randomization takes place (i.e. for each yearchamber-cohort-caucus district group). The variable relative rank is the scaled random seniority
rank of each legislator within their randomization group. The variable ranges from 0 to 1, with
legislators assigned to 1 as the most senior. Coefficients represent the estimated effects of being
the most senior member instead of the least senior member. Ns differ in regressions with
dependent variables for which data is not available for all years.
41
Table A3. Effect of Seniority within Cohort (Relative Rank) on Outcomes of Interest, by Legislator Tenure (OLS)
Win Reelection
Lose Primary
Lose General
Run for Higher Office
Win Higher Office
F.E.
No F.E.
F.E.
No F.E.
F.E.
No F.E.
F.E.
No F.E.
F.E.
No F.E.
Rel. Rank
-0.018
-0.001
0.010
0.008
0.019
0.018
0.002
0.002
0.007
0.007
(0.044)
(0.060)
(0.021)
(0.022)
(0.013)
(0.013)
(0.026)
(0.026)
(0.019)
(0.019)
Rank*
-0.033
-0.031
-0.002
0.001
-0.031
-0.031
0.053
0.053
0.043
0.043
2nd Term
(0.064)
(0.086)
(0.031)
(0.032)
(0.019)
(0.019)
(0.037)
(0.038)
(0.027)
(0.027)
Rank*
0.066
0.054
-0.015
-0.019
-0.019
-0.018
-0.027
-0.027
-0.013
-0.013
3rd+ Term (0.053)
(0.071)
(0.025)
(0.026)
(0.016)
(0.016)
(0.030)
(0.031)
(0.022)
(0.022)
2nd Term
-0.046
-0.000
0.019
-0.000
-0.004
(0.052)
(0.020)
(0.012)
(0.023)
(0.016)
3rd+ Term
-0.256**
0.026
0.004
0.045*
0.016
(0.044)
(0.016)
(0.010)
(0.019)
(0.014)
Rel. Rank
Rank*
2nd Term
Rank*
3rd+ Term
nd
2 Term
3rd+ Term
Opposed in General
F.E.
No F.E.
0.007
0.001
(0.041)
(0.042)
0.032
0.027
(0.060)
(0.061)
0.002
0.007
(0.049)
(0.050)
-0.069^
(0.037)
-0.099**
(0.031)
Opposed in Primary
F.E.
No F.E.
-0.025
-0.022
(0.042)
(0.044)
-0.012
-0.005
(0.060)
(0.063)
0.061
0.046
(0.049)
(0.052)
0.011
(0.039)
0.006
(0.032)
Vote Share in General
F.E.
No F.E.
-0.054
-0.038
(0.045)
(0.037)
0.121^
0.063
(0.068)
(0.053)
-0.019
0.005
(0.089)
(0.048)
-0.035
(0.032)
0.029
(0.030)
Vote Share in Primary
F.E.
No F.E.
-0.125
-0.088^
(0.078)
(0.050)
0.093
0.050
(0.115)
(0.072)
0.143
0.115*
(0.092)
(0.057)
-0.029
(0.041)
-0.10**
(0.034)
Retire
F.E.
No F.E.
-0.019
-0.030
(0.038) (0.047)
0.019
0.021
(0.055) (0.068)
0.006
0.013
(0.045) (0.056)
0.029
(0.042)
0.229**
(0.035)
Money Raised
F.E.
No F.E.
-11,228
-7,409
(8,360)
(8,458)
5,611
8,432
(12,336)
(12,458)
-235,210^
-48,337
(132,506)
(62,987)
-3,710
(7,501.557)
103,791*
(40,460)
Serve Chamber Leader
Number Bills Filed
Number Bills Passed
Party Unity
Unity (Losing Votes)
F.E.
No F.E.
F.E.
No F.E.
F.E.
No F.E.
F.E.
No F.E.
F.E.
No F.E.
Rel. Rank
-0.013
-0.013
0.668
0.838
-0.446
-0.340
0.044
0.044
0.012
0.012
(0.027)
(0.027)
(3.197)
(3.642)
(2.190)
(2.567)
(0.055)
(0.061)
(0.022)
(0.024)
Rank*
-0.041
-0.041
1.117
1.499
2.991
3.217
-0.053
-0.054
-0.006
-0.007
2nd Term
(0.039)
(0.039)
(4.461)
(5.083)
(3.057)
(3.582)
(0.080)
(0.088)
(0.031)
(0.034)
Rank*
0.032
0.032
-11.5*
-11.2*
-3.874
-3.701
-0.101
-0.100
-0.032
-0.032
3rd+ Term (0.032)
(0.032)
(4.5)
(5.1)
(3.062)
(3.588)
(0.079)
(0.086)
(0.031)
(0.034)
2nd Term
0.073**
6.7*
2.624
0.043
0.019
(0.024)
(3.0)
(2.130)
(0.052)
(0.020)
3rd+ Term
0.039^
13.7**
6.604**
0.068
0.024
(0.020)
(3.024)
(2.131)
(0.052)
(0.020)
Notes: Fixed effects for year-chamber-cohort-caucus district groups. See paper for variable description. Sig. Levels: ^0.10, *0.05, **0.01.
Extremity
F.E.
No F.E.
0.005
0.008
(0.037) (0.040)
0.018
0.022
(0.054) (0.058)
-0.061
-0.069
(0.053) (0.057)
-0.015
(0.035)
0.010
(0.035)
42
Table A4. Effect of Seniority within Cohort (Relative Rank) on Outcomes of Interest (OLS)
Win Reelection
Lose Primary
Lose General
Run for Higher Office
F.E.
No F.E.
F.E.
No F.E.
F.E.
No F.E.
F.E.
No F.E.
Rel. Rank
0.002
0.007
-0.004
-0.007
0.003
0.003
-0.001
-0.001
(0.032)
(0.043)
(0.015)
(0.016)
(0.010)
(0.010)
(0.018)
(0.019)
Rank*90s
0.070
0.070
-0.001
-0.004
-0.006
-0.005
-0.024
-0.024
(0.051)
(0.068)
(0.025)
(0.026)
(0.016)
(0.016)
(0.030)
(0.030)
Rank*00s
-0.028
0.018
0.010
0.015
-0.002
-0.003
0.018
0.018
(0.050)
(0.066)
(0.024)
(0.025)
(0.015)
(0.015)
(0.029)
(0.029)
1990s
-0.170**
-0.031^
0.006
0.059**
(0.043)
(0.016)
(0.010)
(0.019)
2000s
-0.332**
-0.052**
0.003
0.048**
(0.041)
(0.015)
(0.009)
(0.018)
Rel. Rank
Rank*90s
Rank*00s
1990s
2000s
Opposed in General
F.E.
No F.E.
0.009
0.012
(0.030)
(0.030)
0.002
-0.014
(0.048)
(0.048)
0.014
0.009
(0.046)
(0.047)
0.071*
(0.031)
-0.029
(0.029)
Opposed in Primary
F.E.
No F.E.
-0.008
-0.013
(0.030)
(0.031)
0.062
0.053
(0.048)
(0.050)
-0.014
0.003
(0.046)
(0.048)
-0.097**
(0.032)
-0.135**
(0.030)
Vote Share in General
F.E.
No F.E.
-0.062
-0.051^
(0.061)
(0.031)
0.069
0.049
(0.077)
(0.044)
0.065
0.078
(0.085)
(0.053)
-0.067*
(0.028)
-0.084*
(0.033)
Vote Share in Primary
F.E.
No F.E.
-0.047
-0.005
(0.050)
(0.028)
0.047
0.013
(0.082)
(0.049)
0.077
-0.090
(0.151)
(0.083)
0.021
(0.032)
0.054
(0.049)
Win Higher Office
F.E.
No F.E.
0.007
0.007
(0.013)
(0.013)
-0.007
-0.007
(0.021)
(0.022)
0.011
0.011
(0.021)
(0.021)
0.028^
(0.014)
0.018
(0.013)
Retire
F.E.
No F.E.
0.001
-0.001
(0.027)
(0.035)
-0.029
-0.035
(0.044)
(0.056)
-0.007
-0.029
(0.042)
(0.055)
0.129**
(0.036)
0.142**
(0.034)
Money Raised
F.E.
No F.E.
-6,054
-2,947
(14,736)
(6,917)
0.000
-2,500
(0.000)
(16,879)
-3,768
0.000
(16,221)
(0.000)
-11,293
(10,030)
0.000
(0.000)
Serve Chamber Leader
Number Bills Filed
Number Bills Passed
Party Unity
Unity (Losing Votes)
F.E.
No F.E.
F.E.
No F.E.
F.E.
No F.E.
F.E.
No F.E.
F.E.
No F.E.
Rel. Rank
0.017
0.017
-2.831
-2.543
-0.736
-0.558
-0.057
0.013
0.002
-0.000
(0.019)
(0.019)
(1.845)
(2.194)
(1.252)
(1.511)
(0.063)
(0.040)
(0.014)
(0.027)
Rank*90s
-0.022
-0.022
0.000
0.000
0.000
0.000
0.000
-0.070
-0.002
0.000
(0.031)
(0.031)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.080)
(0.029)
(0.000)
Rank*00s
-0.041
-0.041
0.000
0.000
0.000
0.000
0.070
0.000
0.000
0.002
(0.030)
(0.030)
(0.000)
(0.000)
(0.000)
(0.000)
(0.073)
(0.000)
(0.000)
(0.031)
1990s
0.030
0.000
0.000
0.066
-0.056**
(0.020)
(0.000)
(0.000)
(0.049)
(0.019)
2000s
0.100**
0.000
0.000
0.000
0.000
(0.019)
(0.000)
(0.000)
(0.000)
(0.000)
Notes: Fixed effects for year-chamber-cohort-caucus district groups. See paper for variable description. Sig. Levels: ^0.10, *0.05, **0.01.
Extremity
F.E.
No F.E.
-0.076^ -0.077^
(0.042)
(0.045)
0.000
0.000
(0.000)
(0.000)
0.097*
0.101^
(0.049)
(0.052)
-0.070*
(0.032)
0.000
(0.000)