Saïd Business School
Research Papers
November 2016
Revealing the Anatomy of Vote Trading
Dr Omar A. Guerrero
Saïd Business School, University of Oxford
Dr Ulrich Matter
Berkman Klein Center for Internet & Society, Harvard University
Faculty of Business and Economics, University of Basel
Saïd Business School RP 2016-26
The Saïd Business School’s working paper series aims to provide early access to high-quality and rigorous academic research. Oxford Saïd’s
working papers reflect a commitment to excellence, and an interdisciplinary scope that is appropriate to a business school embedded in one of the
world’s major research universities..
This paper is authorised or co-authored by Oxford Saïd faculty. It is circulated for comment and discussion only. Contents should be considered
preliminary, and are not to be quoted or reproduced without the author’s permission.
Revealing the Anatomy of Vote Trading
Omar A. Guerrero,1∗ Ulrich Matter,2
1
CABDyN Complexity Centre, Saı̈d Business School, University of Oxford.
2
Berkman Klein Center for Internet & Society, Harvard University;
Faculty of Business and Economics, University of Basel.
∗
E-mail: [email protected]
Supplementary material available upon request.
Cooperation in the form of vote trading, also known as logrolling, is central for
law-making processes, shaping the development of democratic societies. Empirical evidence of logrolling is scarce and limited to highly specific situations
because existing methods are not easily applicable to broader contexts. We
have developed a general and scalable methodology for revealing a network
of vote traders, allowing us to measure logrolling on a large scale. Analysis
on more than 9 million votes spanning 40 years in the U.S. Congress reveals
a higher logrolling prevalence in the Senate and an overall decreasing trend
over recent congresses, coincidental with high levels of political polarization.
Our method is applicable in multiple contexts, shedding light on many aspects
of logrolling and opening new doors in the study of hidden cooperation.
1
Introduction
The law-making process of any democratic society is critical to its socioeconomic development. For example, in 2013, a political gridlock in the U.S. Congress led to a government
shutdown with an estimated economic cost of at least 0.3% of the quarterly GDP (1). In this
case, cooperation between legislators with opposing political views was necessary to end this
blockage. However, in other situations, this kind of cooperation might just as well favor special
interest groups and thereby hurt third parties. Here, we look at vote trading, or logrolling, as
one such type of cooperative behavior in which two legislators vote in favor of each others’
preferred bills in order to secure their passage. While it is extensively discussed in theoretical literature from Economics, Game Theory, and Political Science (2–7), empirical evidence
on vote trading is scarce and limited to the study of a few specific votes or legislators. Up
to now we have lacked the tools with which to study the prevalence of logrolling on a large
scale. Insights into the relative importance of this hidden cooperation in legislative assemblies,
however, are crucial in order to understand the success and failure of alternative institutional
arrangements used to address economic and social problems. Leaving aside value judgements
about the benefits, costs, and ethics of this type of cooperation, all social sciences dedicated to
the study of politics consider vote trading – on both theoretical grounds and anecdotal evidence
– a central aspect of democratic processes (not only in legislative assemblies, but also in any
group making decisions with the simple majority rule). However, without a widely applicable
general approach that allows vote trading to be captured empirically, our understanding of its
determinants, mechanisms, and consequences is limited.
The challenge of empirically assessing the prevalence of logrolling lies in its nature. Legislators on both sides of a vote-trading agreement are keen to keep such a deal secret. Strategically
offering one’s vote to a fellow legislator with opposing political views can contrast starkly with
2
the expectation of voters and peers with similar political views. Hence, empirically capturing
vote trading requires us to measure something that is not directly observable. Previous empirical studies that have tried to detect or measure vote trading focused on a few specific bills (8, 9)
or covered trades within a small group of legislators (10). These approaches are rather designed
for and very useful in analyzing highly specific aspects of vote trading. Because they require
extensive knowledge about the sociopolitical context of specific bills and legislators, they are
unfit for measuring logrolling on a large scale. We have developed a novel approach to measuring the prevalence of vote trading by combining new insights from complex networks literature
with computational methods and well-established theory on logrolling.
When seen as a systemic phenomenon, vote trading is driven by fundamental incentives
set by a common democratic procedure: voting under the simple majority rule. As legislators
have only one vote per decision, this rule cannot take into account preference intensities for
different bills (that is, legislators cannot weight their vote on a bill depending on whether they
particularly favor or dislike it). They simply have one vote per issue. However, legislators
can give up their preferred position on a bill and ‘offer’ their vote in order to get support in
the passage of their favorite bills. Vote trading is thus mutually beneficial for legislators who
are particularly interested in the passage of specific bills. It also implies that when trading
their votes, legislators do not vote sincerely, i.e., in line with their preferred policy position.
Importantly, offering to vote in favor of bills one dislikes in order to get support for one’s
most favored bill(s) primarily makes sense in the context of narrow voting outcomes (8, 10).
In narrow outcomes, the marginal support is particularly valuable for the supporters of the bill.
With higher incentives to trade votes, roll calls in narrow outcomes do not necessarily reveal
legislators’ actual political positions on an issue but could be the result of exchanging favors.
Conversely, legislators’ observed decisions in roll calls decided by a wide margin are more
likely to reveal their preferred policy positions (with a few exceptions such as protest voting
3
(11)). In roll calls, legislators engaged in vote trading can observe each others’ actions once
the votes are cast and thus know with certainty whether their partner kept his or her part of the
bargain. However, as they have strong incentives to keep such deals secret, they cannot directly
punish defection (e.g., publicly blame or sue a legislator who does not keep his or her part of
the bargain). Cooperation in the form of logrolling is thus only likely to evolve as reciprocal
behavior over repeated interactions (12). With this framework in mind, we have developed a
methodology that measures logrolling on large-scale roll call datasets. The theoretical pillars
of the method can be summarized in four principles that have been tested in specific voting
contexts (8, 9):
1. Incentives to trade votes are stronger the narrower vote outcomes are.
2. If a legislator trades a vote, then he or she votes in the opposite direction to what would
be predicted, i.e. the legislator deviates from his or her preferred position.
3. A deviation is considered a potential trade if it is directed to benefit a legislator with a
clear interest in passing the bill that is being voted on.
4. Directed deviations are considered traded votes only if they are reciprocal and mutually
profitable.
In a nutshell, our method consists of four components: estimating legislators’ preferred
policy positions; detecting deviations from those positions; identifying the beneficiaries of those
deviations; and measuring reciprocity between deviators and beneficiaries. The result is an
index that measures the prevalence of vote trading, and a network of vote traders. Its modular
design provides great flexibility so that researchers can adapt it to other contexts by modifying
these components.
4
Methodology
We analyze a dataset covering more than 9 million votes over a period of 40 years of roll calls
and bill (co)sponsorships in the U.S. Congress. Voting data comes from official roll call records
published on the websites of the U.S. Senate and House (see Supplementary Material for a
detailed description of the data as well as the data sources). For each chamber, we encode
voting information into a matrix V. Let entry Vik = 1 if legislator i voted Yes in roll call k and
Vik = 0 if he or she voted No. V has dimension N × K, where N is the number of legislators
and K the number of roll calls.
In line with (8, 10) we assume that legislators’ perceptions about the narrowness of vote
outcomes can be approximated by the realized outcomes. Thus the narrower an observed vote
outcome, the more likely is the presence of incentives to trade votes. However, narrow vote
outcomes are rare and legislators arranging a trade of votes might prefer to arrange coalitions
greater than the minimal winning coalition due to uncertainty as to other legislators’ support
(13, 14). Choosing the margin is thus a trade-off between the size of the set of potentially
traded votes and the inclusion of only those votes posing strong incentives to trade. We define
a narrow margin to be five or fewer votes and consider wide margin outcomes those that were
passed or turned down with at least 20% of the votes (see Supplementary Material for further
details and robustness). In such clear outcomes we assume that legislators are likely to vote
truthfully according to their preferred positions. We thus infer their preferred positions, or ideal
points, as follows.
Consider a one-dimensional policy space where each bill can be located. Each legislator
has an ideal point in this space, and so does each Yes or No vote on a particular bill (higher
dimensional spaces can also be used if necessary, however, it has been shown that the first
dimension captures over 70% of the variation in voting decisions in the U.S. Congress (15,
5
16)). An important assumption in this kind of spatial voting models is that legislators suffer
disutility the further their vote decision is from their ideal points in this policy space. Since
wide margin roll calls are more likely to elicit sincere voting, we use them to estimate the
ideal points in order to infer their preferred policy positions. For this purpose, we employ the
Bayesian approach suggested by (17) and estimate the ideal point x̂i of every legislator i (see
Supplementary Material for details on this step). Alternative methods (15, 18) can also be used
but it has been shown that (17) differentiates better between extreme points. The ideal points
allow us to compute the probability that legislator i will vote Yes in roll call k. For this, we
estimate for each roll call the probit model Pr(Vik = 1|x̂i ) = Φ(x̂i βk ), where Φ is the standard
normal CDF and βk is the parameter to be estimated. We use this model to predict legislators’
probabilities to vote Yes (see Supplementary Material for robustness to prediction errors). We
collect these predictions in a matrix Q with dimensions N × K. Q is key to detect deviations
from preferred policy positions.
Let τ be a probability threshold to consider if a Yes vote is a deviation. For example, suppose
that legislator i votes Yes in a narrow roll call. However, he or she was predicted to vote No
with 0.9 probability, i.e. Qik = 0.1. Then, i’s vote is considered a deviation only if Qik ≤ τ .
Of course, No votes can also be considered deviations, for example, if a legislator attempts
to block the passage of a bill he or she is predicted to be in favor of. We focus on Yes votes
because their beneficiaries can be deducted from different data (e.g. in (co)sponsorship data,
constituency characteristics, or campaign finance data) and the large majority of theoretical
work concentrates on them.
In order to identify the beneficiary of a deviation, we use (co)sponsorship data (which are
assumed to approximate the legislator’s preferences towards bills) based on the official bill
data published by the Library of Congress. We encode this information into a matrix S with
dimensions N × K, where entry Sik = 1 if legislator i (co)sponsored the bill voted on in roll
6
call k and Sik = 0 otherwise. Therefore, i’s vote in a narrow roll call k is considered a directed
deviation only if Vik = 1, Qik ≤ τ , and Sjk = 1 for some j 6= i. For our analysis, we
assume that (co)sponsorships capture the explicit interests of legislators towards the passage of
bills. However, other factors such as social ties or strong interests among the representatives’
constituencies toward the bills could be encoded in S as well.
Now we can construct the directed deviation network (DDN). In the DDN, an edge i → j
indicates that legislator i deviates in a roll call associated to a bill (co)sponsored by legislator
j. The DDN is represented by the adjacency matrix W, where entry Wij > 0 if i deviates
one or more times to the benefit of j, and Wij = 0 otherwise. We measure the reciprocity
in W by adapting the method proposed by (19) to the context of logrolling. When defining
reciprocity between legislators i and j as the symmetric part of their connections, expressed by
↔
↔
wij
= min [Wij , Wji ] = wji
, the function
(
P
↔
1 if N
j6=i wij > 0
fi (W) =
0 otherwise
indicates whether legislator i engaged in at least one reciprocal deviation. Then, the total numP
ber of representatives that reciprocate deviations is R = N
i fi (W) and the total number of
P PN
deviations is W = N
i
j6=i Wij . The level of vote trading in the DDN is thus measured by
t=
2R
,
W
where 2R is a suitable predictor of the true number of votes traded. The total number of
reciprocal edges is usually employed to measure network reciprocity (19–24). In our context,
however, we can show (based on a Monte Carlo Study) that the usually applied predictor inflates
t because it is significantly larger than the true number of trades. 2R is generally lower, so it
corrects this bias (see SI for more details on the Monte Carlo simulation where we can control
for each true trade).
The magnitude of t might be driven by the topology of the DDN and thus by the data at
hand. Therefore, we test the alternative hypothesis that deviations are the result of random
7
errors rather than intentional behavior to benefit other legislators. For this test, we generate a
sample of DDNs under the null model of random deviations. That is, we model each deviation
as an independent Bernoulli random variable with probability of success Qik . By simulating
all the observed deviations in V, we generate a null DDN. Repeating this procedure multiple
times yields a sample of DDNs. We calculate the level of vote trading t0 of each null DDN and
compute the sample mean t̄0 , which reflects the expected trading level under the null hypothesis.
Following (19, 24), we construct the logrolling index
`=
t − t̄0
.
1 − t̄0
(1)
The sign of ` is directly informative of the reciprocity in the DDN. If ` > 0, it means that
legislators tend to deviate in favor of other legislators in a reciprocal fashion, suggesting vote
trading. If ` < 0, then legislators tend to avoid reciprocity, indicating that the directed deviations
are caused by other mechanisms.
The final step is constructing the vote-trading network (VTN). For this, we extract the reciprocal part of the DDN and remove pairs of reciprocal directed deviations that are not mutually
profitable. Recall that the spatial model implies decreasing utility, and hence decreasing probability, of a Yes vote as the distance between the Yes position and the ideal points grows. Let
ki and kj denote roll calls associated to bills that are (co)sponsored by i and j respectively. In
addition, assume that i deviated in kj and j deviated in ki . Profitability means that i’s Yes vote
in kj yields more utility than his or her Yes vote in ki . Therefore, trades are mutually profitable
only if Qiki < Qikj and Qjkj < Qjki .
Results
Figure 1 shows the result of estimating (1) for different values of τ in the two chambers. In
both cases, ` is positive and statistically significant for lower values of τ . This result is intuitive
8
because it suggests that vote trading can only be detected when the probability of a Yes is lower
than that of a No. In other words, if we would expect legislators to vote Yes most of the time,
then the DDN would be full of false deviations. Since we are interested in gaining as much
information as possible about the VTN of each chamber, we choose for the following analyses
the largest possible value of τ such that ` is positive and statistically significant. Figure 2 shows
the VTNs of the Senate and the House. The graphs reveal that, while trades among members of
the same party are common, a large part of vote trading is bipartisan. This is true for both the
Senate and the House.
Table 1 shows the summary statistics of the VTNs from the Senate and the House with respect to party affiliation and network characteristics. In addition, the table lists the congressmen
with most trades in the VTNs as well as the top coalitions (see Supplementary Material for the
complete lists). These results are in line with qualitative studies and theoretical work on vote
trading. For example, it has been argued that Senators are relatively more engaged in vote trading than Representatives and, more generally, that longer tenureships in office might be a central
factor for stable vote-trading relationships (6, 25). The average level of the logrolling index in
its positive region is significantly higher in the Senate. Relatively more Senators participated in
trades than Representatives, and relatively more roll calls (and more bills) were affected in the
Senate than in the House. In addition, the average number of trades per dyad is higher in the
Senate, suggesting more intensity in vote-trading activity. Lengthier time between trades in the
Senate is consistent with the idea that longer relationships between Senators (fostered by the
institution of 6-year terms) are important in building up trust in order to effectively engage in
vote trading.
Figure 3 depicts how logrolling in the Senate and House has evolved over the last four
decades with respect to two potentially important drivers of vote trading: party division and political polarization (15). The figure suggests that vote trading is common in most congresses but
9
varies substantially over time. Within-party trading is particularly present in, but not restricted
to, the majority party. Interestingly, bipartisan cooperation in the form of vote trading is a common feature of the political process in both chambers during most congresses. Moreover, the
share of bipartisan trades is substantially higher in the Senate compared to the House. This finding is in line with the widespread view that party politics is less relevant in the Senate compared
to the House (26). In recent congresses (with very high levels of polarization) both partisan
and bipartisan trading has strongly declined. Note that the decrease in Republican trades stands
in stark contrast to previous congresses with Republican majorities, indicating that inner-party
struggles related to the advent of the Tea Party Caucus within the Republican Party might have
substantially hindered within-party cooperation (27).
The topology of the VTN is informative about the concentration of trades among legislators. The left panel in Figure 4 presents both degree distributions. Interestingly, both chambers
exhibit a Gamma distribution
Pr(x; α, β) =
x
1
α−1 − β
.
x
e
Γ(α)β α
(2)
It is remarkable that the degrees of these independent VTNs are well explained by this distribution. Furthermore, the null hypothesis fails to generate VTNs with Gamma-distributed degrees
(see Supplementary Material), confirming that these findings are a feature of the data rather than
an artifact of our methodology.
These results can be explained by an underlying random variable X, distributed across
legislators according to a normal N (0, σ) (right panel in Figure 4). X may represent a factor
that causes some legislators to be more successful than others at accomplishing successful vote
trades. We represent this mechanism by Y = X 2 , where Y is the degree. Then, by the properties
of sums of squared normal random variables, Y ∼ Gamma 21 , 2σ 2 . Aderson-Darling tests fail
to reject that the empirical data comes from this distribution (see Figure 4).
10
This model can be used to test candidate explanations of the distribution of trades across
legislators, for example, social skills (28). It can also be extended to investigate logrolling from
a network perspective, taking into account contextual knowledge about the various motivations
why specific legislators would choose to trade with each other. The specifics about who connects with whom can be modeled through, for example, different variations of the configuration
model (29), or agent-computing models (30, 31).
Conclusions
The method presented combines well-established theory from Political Science, Economics,
and Game Theory with big data and network science (both of which are considered increasingly
relevant for new scientific discoveries in Political Science and Economics (32–36)). However,
it has some limitations that could be addressed in future research. Although we can test the
statistical significance of logrolling at the aggregate level, we cannot tell whether all the reciprocal deviations revealed in the VTN are true trades (hence we use 2R to reduce false positives).
Any conclusions with respect to low-degree legislators in the VTN thus need to be considered
with care. On a more general level, it is important to keep in mind that there might well be
other forms of vote trading. Three aspects are of particular relevance in this context. First,
(co)sponsorships are not the only signal of preferences towards bills. Second, it might well be
that some deals are arranged between groups. In those cases, our method would likely capture the roll call/bill affected by such trades, but not each individual participating. Third, our
method does not capture ‘implicit’ vote trading (i.e., trades in the form of issue-packages or
‘platforms’ (4) and exchanging favors in order to support or oppose policies before they can
officially be voted on).
Previous research has taken an econometric approach (like the seminal work by (8)) that is
hardly scalable and relies on ex-ante knowing the specific socioeconomic context of bills and
11
legislators where vote trading is suspected. In contrast, our method is designed to study general
aspects of logrolling on a large scale. The approach is flexible, scalable, and provides results
in a very intuitive and informative manner (e.g. through VTNs), making it easily applicable
to various legislative assemblies and time frames. In fact, with more voting and sponsorship
data available, our method can be applied to study hidden cooperation in the form of vote
trading across U.S. states, different counties, as well as different types of assemblies. While
far from conclusive, our results present first insights into vote trading as a means to potentially
overcome political obstruction generated by increasing ideological polarization. However, such
hidden cooperation in the presence of increasing polarization is only possible up to a certain
point of ideological cleavage. Insights in this direction are important in informing the design
of democratic institutions. More broadly, the ability to detect and measure hidden cooperation
such as logrolling might reveal important insights about the dynamics underpinning hidden
human cooperation in general.
References and Notes
1. Gross Domestic Product. Fourth Quarter of 2013., Technical note, Bureau of Economic
Analysis, U.S. Department of Commerce (2014).
2. J. Buchanan, G. Tullock, The Calculus of Consent: Logical Foundations of Constitutional
Democracy (University of Michigan Press, Ann Arbor, MI, 1962).
3. R. Wilson, American Economic Review 59, 331 (1969).
4. G. Tullock, American Economic Review 60, 419 (1970).
5. W. Riker, S. J. Brams, American Political Science Review 67, 1235 (1973).
6. P. Bernholz, Public Choice 33, 65 (1978).
12
7. A. Casella, T. Palfrey, S. Turban, Journal of Public Economics 112, 115 (2014).
8. T. Stratmann, American Economic Review 82, 1162 (1992).
9. T. Stratmann, Economic Inquiry 33, 441 (1995).
10. L. Cohen, C. Malloy, American Economic Journal: Economic Policy 6, 63 (2014).
11. S. Asch, Psychological Monographs: General and Applied 70, 1 (1956).
12. R. Axelrod, The Evolution of Cooperation (Basic Books, New York, NY, 1984).
13. B. Weingast, American Journal of Political Science 23, 245 (1979).
14. T. Groseclose, J. Snyder, American Political Science Review 90, 303 (1996).
15. K. Poole, H. Rosenthal, American Journal of Political Science 29, 357 (1985).
16. K. Poole, H. Rosenthal, Ideology & Congress (Transaction Publishers, New Brunswick,
NJ, 2007).
17. J. Clinton, S. Jackman, D. Rivers, American Political Science Review 98, 355 (2004).
18. M. Porter, P. Mucha, M. Newman, C. Warmbrand, PNAS 102, 7057 (2005).
19. T. Squartini, F. Picciolo, F. Ruzzenenti, D. Garlaschelli, Scientific Reports 3 (2013).
20. L. Katz, J. Powell, Sociometry 18, 403 (1955).
21. S. Achuthan, S. Rao, A. Rao, Proceedings of the Seminar on Combinatorics and Applications in Honour of Professor S. S. Shrinkhande on His 65th Birthday, K. Vijayan, N. Singhi,
eds. (Calcutta: Indian Statistical Institute, 1982), pp. 8–20.
22. C. Wang, et al., Network Science 1, 31 (2013).
13
23. L. Akoglu, P. Melo, C. Faloutsos, Advances in Knowledge Discovery and Data Mining,
P. Tan, S. Chawla, C. Ho, J. Bailey, eds., no. 7302 in Lecture Notes in Computer Science
(Springer Berlin Heidelberg, 2012), pp. 85–96.
24. D. Garlaschelli, M. Loffredo, Physical Review Letters 93, 268701 (2004).
25. D. Matthews, U.S. Senators and their World (W. W. Norton & Company, N.Y., 1973).
26. N. Monroe, J. Roberts, D. Rohde, eds., Why Not Parties?: Party Effects in the United States
Senate (University of Chicago Press, Chicago, IL, 2008).
27. G. Kabaservice, Rule and Ruin: The Downfall of Moderation and the Destruction of the
Republican Party, from Eisenhower to the Tea Party (Oxford University Press, USA, New
York, NY, 2013).
28. D. Deming, The growing importance of social skills in the labor market, Working Paper
21473, National Bureau of Economic Research (2015).
29. M. Molloy, B. Reed, Random Structures & Algorithms 6, 161 (1995).
30. R. Axelrod, Complexity of Cooperation: Agent-Based Models of Competition and Collaboration (Princeton University Press, Princeton, NJ, 1997).
31. R. Axtell, Why agents? On the varied motivations for agent computing in the social sciences (2000).
32. J. Fowler, Political Analysis 14, 456 (2006).
33. D. Lazer, et al., Science 323, 721 (2009).
34. L. Einav, J. Levin, Science 346, 715 (2014).
14
35. M. Jackson, Journal of Economic Perspectives 28, 3 (2014).
36. S. Cranmer, E. Menninga, P. Mucha, Proceedings of the National Academy of Sciences 112,
11812 (2015).
37. M. Newman, Physical Review E 67, 026126 (2003).
15
Figure 1: Empirical relationship between τ and `. The index was computed by generating
samples of 10,000 null DDNs for each level of τ . Standard errors of ` were computed using the
Jackknife procedure proposed by (37).
Senate
0.04
index estimation
95% confidence interval
0.03
logrolling index `
logrolling index `
0.04
0.02
0.01
0.00
0.01
0.02
0.03
0.0
0.2
0.4
0.6
probability threshold τ
0.8
1.0
16
House of Representatives
index estimation
95% confidence interval
0.03
0.02
0.01
0.00
0.01
0.02
0.03
0.0
0.2
0.4
0.6
probability threshold τ
0.8
1.0
Figure 2: Vote-trading network of the Senate. Nodes are colored according to the party of
longest affiliation. Their sizes are proportional to their degree. If the color of an edge does not
match the color of its nodes (e.g. blue edge between red nodes), it is because at least one of the
nodes had a party affiliation at the moment of the vote that was different from the one captured
by its color. We recommend this graphic is viewed digitally (rather than in hard copy format)
to allow easy magnification of legislators’ names.
Helms, Jesse (NC)
Hatfield, Ma
Warner, Johnrk (OR)
(VA)
Shelb
Lugar,y, Richard (AL)
R
ic
h
Thu
ard
Mc rmond, J (IN)
GregCain, Johnames (SC)
(AZ
St g, Ju
Ja evens dd (N )
Dolvits, Ja , Ted (A H)
T e, cob K)
M owe Robe (NY
Gra cClu r, John rt (KS )
)
C s re
H ha sle , Ja (TX
We atch fee, J y, Ch mes ( )
ick , O ohn arle ID)
er, rrin (R s (I
Lo (U I) A)
we T)
ll (
CT
)
sel
Mo
La
Jeffords, James (VT)
Reid, Harry (NV)
)
Harkin, Thomas (IAT)
(V
k
Leahy, Patroicrne (RI)
H)
Pell, Claib
)
owardt(O
r (WV
baum, H
Metzen Byrd, Robeniel (NY) )
S
Da
hn (M V)
ihan,
Moyn tennis, JoJohn (W H)
S
(O I)
r,
n
e
efell , Joh l (W )
Rock Glenn Russeliel (HI L)
,
n (A )
gold , Da ell (LA )
Fein kaka , How ohn (NM I)
A
flin x, J ff (M L)
He reau an, Je Carl ol (I
B
, r
gam vin , Ca
Bin Le raun
-B
ey
Senate
)
NE
s ( N) KY)
e
l
I
(
ar (
Ch niel itch D)
el, , Da ell, M ry (S O)
g
Ha oats onn , Lar rge (C OK)
C cC sler eo ry (
M res n, G Hen DE) )
P row on, am ( (MD
B ellm illi arles
B oth, W s, Ch OK)
R athia Don ( J)
M ckles, ford (N
Ni se, Clif il (TX)
Ca mm, Ph n (CO)
Gra pbell, Be (FL)
Cam kins, Paula
Haw , James (OK)
Inhofe Malcolm (WY)
Wallop,
)
Baker, Howard (TN
Cochran, Thad (MS)
Craig, Larry (ID)
Domenici, Pete (NM)
Boschwitz,
Mack, C Rudolph (MN)
Laxalt, Ponnie (FL)
Quay aul (NV)
Crapole, James (IN)
Bro , Micha
Hutcwnback, S el (ID)
amu
Ru hiso
Simdman, Wn, Kay (T el (KS)
X)
p
s
a
V
o
r
F oinov n, Al ren (N
Ashairclot ich, G an (WY H)
)
S c h, eo
G unu roft Du rge
Br arn, nu, J , John ncan ( (OH)
Co ook Ed ohn (M NC
ve e, win (N O) )
rde Ed (U H)
ll, war T)
Pa d
ul (M
(G A)
A)
IL) )
s( H
rle (O )
ha iam (TN
O)
, C ill am
rcy , W illi C) r (M )
Pe axbeck, W hn (N tophe on (AL
S ro , Jo hris fers
B t
f
)
Easond, Cons, Je im (KY n (NM)
B essi ng, J rriso IL)
S ni Ha
r(
)
Bunhmitt, ld, Pete don (NH
Sc tzgera y, Gor I)
Fi phre rt (W
Humsten, Robe amuel (CA)
Ka akawa, S d (TN)
Hay
n, Fre
o
Thomp, sWilliam (VA)
Scott
el (WY)
Enzi, Micha (VA)
Byrd, Harry
Cranston, Alan (CA)
Morgan, Robert (NC)
C
To lela
rri nd
Tso celli , Jos
n , e
Hu
mp Ba gas, Robe ph (
Abo hrey, yh, E Paul rt (NGA)
McC urezk, Hube van ( (MA) J)
lel Jam rt ( IN)
Pryolan, Jo es (SMN)
John r, M hn (A D)
Sass son, Timark (ARR)
Wells er, Jam (SD) )
to
es
Hask ne, Paul (TN)
Eastlan ell, Floyd ((CMN)
d,
O
Sarbane James (MS))
Dodd, Chriss, Paul (MD)
topher (CT)
Schumer, Charles
(NY)
Metcalf, Lee (MT)
ute Clar
n k
Ino berg , Ric
ber uy , F ha
ma e, ra rd
All n, Jo Dani nk ( (IA)
Gor en, Ja seph el (H NJ)
Talm Bump e, Alb mes ( (CT) I)
er
e A
a
Kenn dge, H s, Dalert (TN L)
erm (A )
e
d
y
Zorin , Ed an ( R)
wa G
sk
Nelso y, Edwarrd (MAA) )
n, Ga d (N
y
Baucu lord (W E)
Melche s, Max (MTI)
Feinstein r, John (MT))
,D
Fulbright, Jaianne (CA)
mes (A
Graham, Bob (FLR))
Durkin, John (NH)
Kerrey, Robert (NE)
Williams, Harrison (NJ)
D)
Dorgan, Byron (N
illiam (ME)
Hathaway, W, Rupert (IN)
Hartke illiam (NJ)
)
,W
Bradley yh, Birch (IANL)
Ba , John ( A)
man y (W
Spark on, Henrald (MI) )
O
s
k
c
Ja gle, Doniam (M CA)
Rie n, Will John ( (MI)
)
,
o
e
ingt ney bbi (DE )
Sym Tun w, De seph (WI A)
o
o m C
ben n, J llia a ( IL)
Sta Bide e, Wi rbar lai ( NH)
ir a d (
xm r, B n, A as
Pro Boxe enso hom
v ,T
Ste tyre
Ic n
M
Lie
democrat trade
republican trade
bipartisan trade
17
democrat
republican
independent
Roukema, Marge (NJ)
Bonker, Don (WA)
(TX)
Stenholm, Charles
Hall, Ralph (TX)
(KY)
Mazzoli,, Romano
Howard (CA)
Berman
(FL)
Fuqua, Don
han (NY)
Jonat
(NJ)
Bingham,
rt
, Robe
icelliilton
TorrHam
, Leel (IN)
(IA)
th,, Nea
Smioks
Jack (TX))
Broon,
Glefornnd (CA
(GAA))
Anders
Bishop,, San
Thomas (W
(CA)
ley
on
Fo
Le
V)
)
Panehattall,, Nickley(W(IA
Ra ll, Berk n (OR)
A)
Bede
en, Ro
Wyd
seph (P
(PA)
A))
ydurosth, Jo
a, Jolvhn
GaM
in (CH
oley, Ca ald (O(OK)
on
Doas
D
el
Pe e, Michaiam (IL)A)
Synar,ki, Willhael (W AO))
Lipins , Mic tney (C
(C
LoSwtaryrk,,FPoratriilciliae (NCV)) )
eder , W lan (W
r (PA )
SchroHefllonheran,r,APeote
y (OHL)
Mo aye , T nrles (I N)
Kostm Haell, CHhaaroeldw(T(CAH)))
, h (O A
PFriocrd
s
tt
, Ma me y (M I)A))
ínezant, JaGerrvid (W
(M
Martrafic dds,y, Darians (SDA)
T
tu
S
ObeellyT, Bhomgalase(C(FL)I)
n
(NNJ))
Donchle,, Dourencme (M
DaBs oscho, Lar,wJeJorosetipnh (M(NYR)I)
SmTitraxinleisoh,, Maatrthneawnd ((MNAY) ) )
r y ( S
M b
a
S
h, M, Fearnme ase (M(NJ)A)
Hug ain , B ho en es (P Y)
McGermFraneky, loTr, GJamnaldm (KMAA))
St. DownTaylorioy,, DWoilliaardrd( (C(TX)A) )
F e , w a y MR
Batcilhery, ERd icHh enerpher( (O(VA)R))
Naarkmeaná,lez,, Jo,sPeatmesese(O(L(APA)J) )
M eh nz arlyzio , J Jamnn ul (N C )
L Go E Fa ran r, ori Pa rd (N CA A)
DeMo aves, Cski,erna esry ( (C O)I)
Weoggnjorr, Be, JaHmernbarrad (dM(M
B a ye rk n, a ha ar
K
DwClaxmaer, ,BRicHow
WaBoxardtlpe,
ph o
Ge W
Ramstad, James (MN)
Shuster, Elmer
(PA)
Sensenbren
ner, F. (WI)
Gingrich Newton
(GA)
Boehle , Sherwo
Upton, rt,
Fred (MI) od (NY)
McKinney
, Stewart (CT)
Meyers, Jan
(KS)
Ehle
rs,
on (MI)
Zschau,Vern
Edw
in
Gan
(CA
ske
)
, Greg (IA)
Young, Don
Be
reuter, Do(AK)
Rit
ter, Donalduglas (NE)
Pe
tri
(PA)
omas (W
Go
od,e,ThVir
Ko
gil (VA)I)
lbe
, James
Le
ac
h,
(AZ)
Ja
Ke
me
m
s (IA
Goodp,linJack (N
Sh
g, WilliaY) )
Po
rtaw
(FL) m (PA)
er,, E.
M
Jo
or
hn
el
reen,la, Const(IL)
GGre
Sedgw ance
en oo
(M
d, Jaic
KDasavicis, wTho
mkes(NY) D)
CHasentlryh, ,PJoauhm
n as
(OH(V)A) (PA)
(M
ome,fieMichl ae
MBro
I)
o
ra
C
ld
Hoaemkpbne, Jerr,y Willli(D
E)
a
G
ll
m
(K
st
ra ra , Carr S) (M
oll (S I)
DBreonildlaison, ,WPete
(MI) C)
DSumnitieh,r, D, Ha enryillisr(O
(T H
FWawolf,cFan,NJoickvid
(M(CA X) )
GCilomllinesll, Hrankh(Vn (TI)N )
N
a
,
Hyudsslean, BMicrhris (IA) )
S
e
,
L
a
e
m
Co ith , H Jamnja el (G)
MMicnate, ,SChernryes (Imin (NA)
FDr eacl,Ker, nJohnilvioisto(Ip L) A) Y)
SKcluagnksNatan, (FL(MAher (N
FKoxingasrb,oSr c,oBt ohbanJo(Ghn) ( ) J)
G
H ek , Joton ou t (W(NJ A) ME)
RBarolleans, Gn (P, Jagch, J I) )
CFoloth,rett beceorA) k (Goe (
EMh clKingeey, MTob, yBillk, Hgae (P A) FL)
B
(
r e r
JoYouate lichon , Wark (W NE roldA)
StRainahnsngm, an,, R,oHoiwllia(mFLI)) ) (NJ)
ng ldoon, C. ( Heber ard (P
ela , MTimFL rb t (M (C A)
nd at ot ) ert D A)
, A the hy (V )
rla w (IL A)
n ( (N )
MNJ)
)
House of Representatives
) ) )
(IL (IL A) A
rt rt (C (C
be be ul ana )
Ro Ro, Pa, D NCCA) ) )
el, y, ey er r ( r ( CAOH
cchClolroskbachaltealteome(rt )( MD) A)
i
MM cC ra , Wr, Wll, T ob FL e ( (W
MRohnersge be n, Ran ( sco Y) rge C)
JoHeamptmar, D , Reor (NGesos (N) A)
CPorilletlettPet utt,, Ca (TNt (C )
MBaring,hercger achobesr (NH) NJ)
KNet llenp, nZ, Rarlen (SaDrd (K) A)
BWa amrna, ChJohRichst (eOrt (AP) )
DBoassune,er,ErneRobas (Il (CO L)
TZhimomok,hlinT,hom
anieS) ard (Ie (RI)
IsCtouugkee, fere, nDt (M
dwudinA)
TSachtat, Tinr skei,r,EClaert (P(IN))
LDoerhwneidr, Rteobpheen(OHT) )
SWc alkyeer, St, Sterrvill g(Ue (CO T)
BCuhabko, M, eGeolr (TXp) her(F(CL)
CBoroowhner, CBhilrisictohae(Nl M))
l C
ASrchaysa,kis, M
anuaerd(T(N
X) )
BLiluirjánle, ,MHonwald
n (NY A)
CPoabul, uRgoh,DJounhcaCn) (C
(S O) Y)
MHcH
ugnlites,r,Bob
(N
elJo(C
Jo
hn
y,
InHefle ney, hn (NVN) )
SEwnseeign, ,JoMarkas(I (IL)H) T)
udiner
ahto(O
So
g, Thebom
orris
pher (U
Ew
ycnne,onD, Ch
es (NJ)
Pr
Ca
ter, Jam
anyk(T(NX)Y) J)
ur
Co
rton,r,FrKaich
l (N
Ho
aeS)
ange
M
Gr
,
as
pp
Pa
rts, ,PaGut y(K(NY)
be
)
Ro
es (LA
linari
Mo
, Jam
omas)(MN)
Crery
, Th
Mc
orn
mas (IA
Haged
Tho
(KY)
ns,
rion
Eva
Ma
,
der n, Jo Ann (MO))
Snyerso
(CA
Em
iam
Will
,
mas
Tho
s (TN)
Jame
len,
Quil
rd (IL)
Edwa
gan,
Madi
ers (OH)
Chalm
Wylie, on,
(PA)
John
Peters
(OH)(OH)
Ney, Robert
Clarence
Brown,
Sam (MO)
Graves,
Henry (SC)
Brown,
Biggert, Judy (IL)
Collins, James (TX)
Cooksey,
John
Miller,
Clarence(LA)
(OH)
Bentley,
Helen (MD)
Winn,
Edward
Whale
n,
Charle(KS)
Saxton, H.
(NJ) s (OH)
Gran
dy,
Fred
erick (IA)
Myrick, Sue
(NC
Rail
)
sbac
Thomas (IL)
Gibbons,k,Jam
Mo
es (NV
linari, Susan
Co
yne
, James (PA(NY))
Gu
bert (MD))
Co
x,de,C.Gil
(CA
Hi
llis
Ta
oo)d (IN)
ylor,, Elw
Ch
M
cD
ar
ade, Joseles (NC)
O'Brien
ph (PA)
Ho
, George
Fr
enlt,zeMl,arWjorie
(IL)
Fo
D)
ella, Viilliam(M(M
W
ilssson
to (NY) N)
RLaedthmam,, HToeather
(N
O
m
on
se
(I
d,
,
A)
Willia M)
GMoocIdntDosouh,g (C
m
(N
A
)
H
avid (I M)
Steuaffinlagtttoe, BDob
N)
n
PWeailsorn
n,s, Cli,ffMich(VaeAl)
HHaamrsseha, ,EdJowea(SoCrd
) (FL)(CA)
BLaecvhy,mDearsWchilliardm(IN)
B
m
(O
v
il
u
id
id
s
b
Kir ray , Sp (N t, JoH)
AWshhitkta, Ma,rBriaennceYr ) hn (AR
)
BKuim,bJarookker,kR(IL) (CA(A
) L)
BSaherlbgeney (C, AJohonbert
RLoigu hfatfolisy, ,LRric, hCla)ir (OH(K
) S)
R
Coongesrseloot, oJ uisa(Frd (A(CA)
C
a
s
Finlev able, Ht, Jo mes L) L)
(IA
HDun dleeyland, Baarroldhn (C
ber (KYA) )
HBakulshcoan, , Pa, Jam
DLewilleaer, Rf, KJohnul (ILes ((NY))
NEerleicnkiniss, Jrey, Vicahaernny(TN)) NH)
H
(
r
d
u
SJenostemaboronn, Wry (nC(T (LAMO)
LSteteerkinsttlenn, , Jo illi A)N) )
MWeewiige s, N, Wr, JMa hn am
MCcomillebrers,, Trh, Wiewitlliaomhnrk( (W(IL) (AL)
Ca bes, JoJohomllia on ( (TIN) I)
m
nd t, hn n ( as ( MDN)
les La (WMN(FL WI )
s, rry A) ) ) )
Alf (T
red X)
(C
A)
A) )
t (M(OH
areas
argomV) ) O)
, M Th (N KS (M Y)
ler ss, n e ( rl (N Y) )
ck ne , Jo, Jo Ea an (N SC ) D)
HKeinodrtebr itmz an,ormeteoryd ((KYael ()I WI))
PSkuole t, Nr, P, Fl Jim ich (IAld ((UT
CLeneysenceing,n, Miley arord N) )
PSpeunnpsoe, Wh, Howna (M) (VT
B im yn lic , H le ID rd H)
a
r
h
S
Maroelsohnl, A.L.B(ern(HI)is (OR) Y)
F
NEierdtaer, eCrsP,atsyDenrnyl r(leA s N(N) )
, e
t d
D
T
a
,
O
SManinckiniconhy,rB, Crhilynre( n D(M) )
)
r (S (CO
KAunthhum,eMa, Paim
I)nd (CO
SLcloydcheolln, Ter, Ehdn (M
o O) )
s utt , Jo Raym(M
A
MJoithnrlm
e yersek, berttin (M
Z)
P
)
R
CKoongonvgs, Rn,oMrarris r(At (O
L)
YMoueeahlla, M,oAlbKearenN(F) I) )
Y )
UUdllmramn anlb, ertn(Taldly(M
n (Nel (MA
J)
TGhoureo,sAta,yD, oCa,roMicha(N
lb ne n enry X)
A
MHaalorrintogsktoi,GeHorgaerd(T(MAA) )
dwrge (CL) T)
HMealshonnd,, Eeo
(Fam (C)
G
la
am
o
r,
li (CT
B
MGiliblebonfos,rdS, ,WSilam
(CT)
on iam (M
A)
RGatejch
dens
r, WillJohn (N
CMotoatekleyy,, Jo
hniamY)(N(LJ)A)
ill
Murph
hes, W, Williamd (MA)
ug
ol
on
H
H)
JeDoffenorshue,JaHmares (O
(OR)
antoenn,aueror, dEa(TrlN) (MD)
St
um , Cliff Charles E)
Bl
Allen
Millen, John (NY)
h,
Mc
ug
(N
na
rt
Cava
)
azek,re,RoGebene (NJ
Mr
s (NY)
Magui
nton, Thoma
(IL)
Ma
Gus
,
(DE) )
Savage
per, Thomas
(MN
Car
ld(MS)
Gera
,
Sikorski
, Alphonso(UT)
Espy
Jim
Matheson,
, Eliot (NY)(TX)
Engel
Geren, Preston
(OH)
d, Ted
Stricklan
(IL)
Bruce, Terry
Ken (TX)
Bentsen,Cecil
(HI)
Heftel,
Rodino, Peter (NJ)
H Mc
Lu GinojInty
n u o r
EdEdwa dinetiérsra, Re, M
wa rd F , S ez ub ike
t ,
r
P M ds s, Tary an Luén (N
Ca Pete ickeitca,, DWilhlom, Jolhey (iNs (I(TX)C)
rn rs Kl t, a ia as n L)
Kas ey, C Joonn, eCin, ORwenieml (C(TX(ILY) )
tR en H hrises, ollinonn (V(FLA) )
oe me Paill, topBe ( (F A )
LTuht ommerie, r, RstorB,arohnern ((GMANL) ) )
Steher,as, RTimobe Ed (INPA))
W ed W ob oth rt (AZ))
HaBrearodlff,, Thoilmliamert (yG(I(W
I
kin , EdLes as (M AN) ) )
GaNredzFi,a,rrT,howmatredr ((NOKN) )
Y
c
HMeraffeiai,, RLoucSieamas(C(I(AR)I) )
H
b
n
te
a
D
e
R r l,
(M A
Reayngerisl, , HeDeannnierlt(N(NY)I) )
es, Ch rbe is (M Y
FeigCollBinryaSnilt,veasrtrlesrt(N(VAI)) )
h s, Jo e Y
CoeMlheaend,s,EdCwardishns (T(TXX))
a
)
rd
o
(I
C
L
(O L)
Kraotorma,nA, nthloonyd (W
Pre
vi Ja y (CAAH))
Doryegr, Ll,unFramnkes (C
Min
A))
an sf
Bacetcha, N,orByroonrd (M
(NCD))
(N
Blatnus
an (C
D)
, Jam
m
ik
A
es
,
Reid, John (FL))
Harry (MN)
WilsMonikva
, Ch, Abner(N
V)
les (O(IHL)
SnydHerall, ar
(NY))
Fi ian,, ViJocthn
(AR)
Bath
oyord (IN
rcia, JaFlme
soeld
MeUn
, Jolenes (M
zvins
I))
Spell
(W
,
Ed
A)
wa
maky
)
dysrd(M(IA
Yon,unGla
g, Joh
ra, Xav
(TXD))
CopperBesmcer
iern(CA
ith,
Samuel
Lantos,
(AZ))
Tom (CA
Daniels
, Dominic
Benjami
(NJ))
n, Adamk (IN)
Poag
e,
Willi
am (TX)
Breckinridge, John
(KY)
Eckhardt, Robert (TX)
Tanner, John (TN)
Mitchell, Harry
Hayes, Charles(AZ)
(IL)
Laughlin, Gregory
Roemer, Charles (TX)
(LA)
Ho
ch S
bru W ola
McKenLoencknhittrez, S
C n g, er n, te
We Slomskeelly, Cla,rGeJoamphen
iss ithy, Bar en rg ie ( (N
Ale
, F bc e M
Fer RoxandDinCgarrT,he,oJosreancarae (M(NYS)Y)
Volkrarowlaner, Well, Mildor phis (I(CTD) )
me, Ged, Ja illiJohtnone( (N(PAN) ) )
Y
r
r
B
,
a
o
ld
H
Gu Ro yd ar inmeasm (A(MMI)I) )
AddLipinYaartinei,, R,obAlloeld e(M(N(GYA)R)
G a s ro Fr e n
)
Orotrdbobno,kJi,o Dna, Gaunkrt( (N(FJLO) )
N
iz
s
)
n
,
B Fa , S Ba sep iel (P J)
Ob Prerrzyio, , oVlomrtoonh( (N(YILA) )
Fog eSrkstaicr,e, DRaobicetrorn(C(TTXN) )
t (A A )
Klelicettae,ltonJamevid (N
zka, Tho, Ikes (M CR) ))
S
h
m ON)
N Jonearp, Gera as (M
(P )
Walg Poewrkaks,,HJaPmheilipld (I(WI)A)
s
e
N
re
in
n
(O
L n, s, ry (N K) )
ZabPloenneyv,in, SDaougClaarl (K
Y)
(PAY))
i, CTimonthders(M
Bouch ck
erC, oForepeler,menyt (M
NI)
Durb
Tauzi in, RicderJiicmk (T(WNI) )
)
n,
h
(V
B
ar
W
ia
Eilberggi, Millariamd(L(ILA) )
io (NY)A)
HatchAspig,
n, Jo
ua
Lesh
(PA)
Ever, Cha
(W
rlslesie (G
I)
Baldans, Dav
(INA))
vinid(W
OttinLageFalcuse,, Al
Jo
r,
hn
Rich (NY)I)
Rees, Th
(NY)
Perkins,omarasd (C
A)
Yates, SidCarl (KY)
Cle
me
ne
Hawkins,nt, Robert y(TN
(IL
gustus (CA ))
Dymally,AuMe
rvyn (CA))
man, Jan
BeilensHar
e (CA)
Eckon, Ant
(CA)
Denhon
nisy(OH
Drinart,
)
an, Robe
rt
(MA
Blanchar
d,
(MI))
Hoagland, Jame
Peters (NE)
Scheuer, James
Evans, Lane(NY)
(IL)
Williams, John
(MT)
Richardson, William
(NM)
Rostenkowski, Daniel (IL)
Ford, William (MI)
Pallone, Frank (NJ)
(FL)
Pepper, Claude(MA)
Mavroules, Nicholas
James (NJ)
Howard, James
Hanley,lo, Jerry(NY)
(IL)
Costel
(CA)
, Jim (OR)
Bates
Lesd (OH)
AuCoin,
Davi
))
Mann,, Har
ley (WV
(NH
Staggers
man
)
Nor
rs,
Dalase (MI
D'AmouKildee, hol
(TX))
(CA
Gary
LampsoCon,ndNic
it,, Llo
(TXD))
yd
arane(M
Dolskggi,ett
(TX)
Ban,rbGe
A)
ku
Mi Gree , Lionel (C
(VA)
S)
erlin Wilbur
ie (M
Van De
(TX)
Daerniy,el,Gillel, sp
al Sanams (T(MX)I))
id (TN )
Montgom FrHost,, Jo
av
D
m (WA )
Bonieror, Willmiaan
O
(C
Bon , Nor hyl (MAA))
au (P Y)
Dicirksth, Tiasm, ot
WTsong, MicPhLaeeol (N(IN
, mes (NC) ))
tiJa
MZyeefrseret
in
,
rt
(OHA)
tz
a
Joner, MMary
st ar, John (C(C(IAL)))
K
LancaOBaukrton, ,Philralipnk (O
on , F n X))
Bnunrtunzsioh, Gz,leCnirnote(T(F(ILL))
A Engli gue , Da neths (AZS)
ri
n me s (M C)
RoFdascaeyll, Ky,eJa
(S )
le
r
r
a
e
G ult Ch utl rs (U(CTA)I)
McoNwdryic, k,DBou,gJolahneil (H(KYC))
D Der ns, ebs , Nroll (N A))
V
e
Ow Krromdb,ieChaarrlemsass( (F(PLAA) )
rc ar C o le t P )
AbHeubRbosell,o, ,TChhRaorbliearmin( ((IOLKV))
riennettski,, WilMart est (WOHV) )
r
e
P Be Boryne so, s, Wber ld ( (W X) )
CoRusatkein, RRoonarleyes ((TW(WY IL) )
WWisttl, , HaJamenoim (F H)A)
O
Mgogerrigs htl,io,odTy,ilJliarmcyp(h ((PCAF)L) )
a
A
Sta WRonc Meoll, Wr, MJaoswearidame( (G(CA)Y)
p u r, l g (N
ap pt lte Ed il or eo s
Ch KaKoybala, n, W, Gean, lLphu
Roehmrden R,yEdo
L Da ns
w
To
democrat trade
republican trade
bipartisan trade
18
democrat
republican
independent
Figure 3: Vote trading, party division, and polarization in the 93rd to the 113th U.S. Congress.
The top panel shows the number of trades between members of the same party as well as bipartisan trades. The middle panel shows the shares of Republican and Democratic seats, with
a horizontal line indicating 50% of the seats. The bottom panel depicts political polarization
in the respective chamber (based on Voteview.com’s broadly applied polarization measure (15),
normalized to the value of the 113th Congress). The dotted orange vertical line indicates the
congress during which the Tea Party movement evolved.
Polarization
Share of seats
No. of trades
Senate
Trade partners
125
Democrats
Bipartisan
Republicans
100
75
50
25
0
0.70
0.65
0.60
0.55
0.50
0.45
0.40
0.35
0.30
1.00
0.90
0.80
0.70
0.60
0.50
Party
Democrats
Republicans
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113
19
Polarization
Share of seats
No. of trades
House of Representatives
Trade partners
500
Democrats
Bipartisan
Republicans
400
300
200
100
0
0.70
0.65
0.60
0.55
0.50
0.45
0.40
0.35
0.30
1.00
0.90
0.80
0.70
0.60
0.50
Party
Democrats
Republicans
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113
20
Figure 4: The VTNs were treated as undirected unweighted graphs because all trades are reciprocal and weights are negligible (see Table 1). We fitted the data by fixing the theoretical shape
parameter α = 0.5 and estimated β = 10.6 for the Senate and β = 14.8 for the House via MLE.
We generated synthetic data from these distributions and performed two-sample AndersonDarling tests. The theoretical model cannot be rejected in any chamber.
Senate σ = 2. 30
House σ = 2. 72
0.20
10 2
10 1
10 0
Attribute Distribution
0.25
density
frequency (CCDF)
VTN Degree Distribution
Senate data
House data
Theoretical
10 0
0.15
0.10
0.05
0.00
10 1
degree
21
10
5
0
X
5
10
Table 1: VTN summary statistics
Attribute
Senate
House
Average reciprocity
0.019 ± 0.008
0.008 ± 0.002
Legislators
140 (37.14%)
538 (31.17%)
By party: Dem. (Rep.) [Ind.]
71 (68) [1]
320 (217) [1]
Bills
173 (5.97%)
228 (2.13%)
Votes
810 (0.07%)
4,114 (0.05%)
Average degree
11.57
15.29
Average trading partners
5.27
7.41
Highest degree (partners)
74 (32)
136 (67)
Average trades per partnership
2.20
2.07
Average years between trades
4.78
4.21
Most central party (votes traded) Democrat (416) Democrat (2,737)
Democrat trades
157
1,874
Republican trades
143
522
Bipartisan trades
510
1,718
Average reciprocity is the mean of ` in its positive and statistically significant region. We
report its average 95% confidence interval. The average number of years between trades was
calculated by averaging the difference between the dates of all possible exchanges of each
dyad from the DDN. This was necessary because in reality we do not know which precise vote
was traded for which other.
22