Professional Documents
Culture Documents
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
American Political Science Association is collaborating with JSTOR to digitize, preserve and extend access to
The American Political Science Review.
http://www.jstor.org
inCongress
vs.DyadicRepresentation
Collective
ROBERTWEISSBERG
University of Illinois, Urbana-Champaign
Previous studies of legislative-constituency representation have focused almost exclusively on
pairs of Congress members and their constituencies. It is possible, however, to think of
representation collectively, i.e., to consider the extent to which Congress as an institution
represents the American people. Our analysis delineates this concept of representation, analyzes its
existence by use of probability theory and the Miller-Stokes data, and then considers the
relationship between collective representation and electoral control. We conclude that citizens
probably get better representation than is suggested by the Miller-Stokes analysis, that the amount
of representation may be more a function of institutional arrangements than of electoral control,
and that citizen indifference towards many aspects of legislative politics is quite reasonable, given
the existence of collective representation.
representing
a people.3
institutions
collectively
535
536
Vol. 72
1978
537
538
According to the dyadic model, the maximum possible degree of representation would
occur if legislators followed the preferences of
their constituents (somehow defined). It would
therefore seem to follow that citizens would
not be represented if (a) all members of
Congress voted randomly or (b) all members of
Congress-for any number of reasons-violated
constituency preferences. We shall argue here
that misrepresentation need not occur if Congress members voted randomly or otherwise
chose to ignore constituency opinion. What
makes such an argument possible is an acceptance of a collective perspective on representation which focuses on the representation of
an opinion within an institution independent of
an electoral relationship between opinion holder and legislator. Let us begin by considering
random legislative voting and collective representation. Subsequently we shall examine how
misrepresentation by some legislators can result
in better overall responsiveness to citizen
preferences.
Random Voting and Representation. As we
have stated, analyses of representation based on
the dyadic model measure representation in
terms of nonrandom association between pairs
of constituencies and legislators (i.e., a high
correlation indicates close representation). Are
we then to equate a zero correlation (i.e.,
perfect randomness) with zero representation?
Our answer is that even with perfectly random
legislative voting it is unlikely that national
opinion would be violated about more than half
the time. In a sense, the use of measures of
association with random baselines, together
with the dyadic approach, can readily lead to
an underestimation of how often majority
preferences are heeded.10 To assess the impact
of perfect randomness, let us assume a legislature of 20 members and constituencies. (We
shall assume an N of 20 since probability
calculations of the type we are going to
perform for an N of 435 would be impossible
Vol. 72
1978
?-
.0000095
.00001907
.00018120
.00108719
.00462006
.01478577
.03696442
.07392883
.12013435
.16017914
.176198001
.16017914
.12013435
.07392883
.03696442
.01478577
.00462006
.00108719
.00018120
.00001907
.0000095
PfrJ=
N!
prq(N-r)
N1!
r! (N-r)
where
N = total number of pairs (20 in this example)
r = number of successes out of 20
(N-r) = number of failures out of 20
P = probability of success (here, .5)
q = probability of failure (here, .5)
539
Nl!(Nl-rl)!
. plrlqNl 1-r.
N2!(N2-r2)!
p2r2q2 N2-r2
Subscripts refer to samples (i.e., 1 = members of
Congress, 2 = constituents); the notation is described
in Table 1.
540
Vol. 72
The AmericanPoliticalScienceReview
Opinion/Roll Call
Scores for
Constituency
Preferences (X)
Legislators
(scale scores)
Differences in
Preferences
District
2
3 Districts
Opinions/Roll Call
Scores for
Constituency
Preferences (X)
Legislators
(scale scores)
Differences in
Preferences
District
2
All
Districts
-2
541
1978
d -
2R
where
d = the mean district scale score
R = the representatives scale score
N = the number of pairs
In words, this formula can be stated thus: the
average legislator-constituency difference on a
dyadic basis is equal to or greater than the
average difference between all legislators and all
constituents. Moreover, for reasons that will be
more fully discussed below, it will usually be
LI d -R I
Id - 2R
so citi>
the case that
N
zens as a whole are better represented by
Congress than are citizens in each district by
their particular legislators.
This principle can be illustrated with the
Miller-Stokes data. Though an ideal illustration
would require that both legislators and constituents be measured on the same scale, we shall
make do with scales composed of different
items and of different ranges.13 Table 3 shows
both dyadic and aggregate (or virtual) representation on the issues relating to social wel13That congressional scores may range 0 to 9 while
constituency scores range from 0 to 3 is no problem
with a correlation coefficient that "standardizes"
these scores. We cannot, however, use standard scores
since these would (by definition) yield means of zero
in column 2 of Table 3, thus rendering our analysis
statistically nonsensical. We should also acknowledge
that the mean or any other measure of central
tendency can be a poor indicator of what is to be
represented, given the wide variety of opinion distribu-
tion in a district. It certainly seems politically important whether or not the distribution is bell-shaped
or U-shaped. These and several other questions dealing
with the measurement of constituency are considered
in greater detail in Robert Weissberg, "Some Issues in
the Analysis of Legislator Constituency Policy Agreement," unpublished mimeographed paper.
140ur use of legislators' party identifiers as "constituency opinion" (as opposed to, say, the entire
district) derives from two considerations. First, of all
the groups we could have used, those belonging to a
Congress member's party seem the most relevant in
terms of representational relationship. Certainly a
member of Congress could not be expected to
represent everyone in the district or partisans of the
opposition party. In any case, the use of the score for
the entire district would not change the thrust of our
analysis. Second, subsequent analysis considers the
theoretically important questions of district minority
representation and this requires that we disaggregate
district opinions.
Issue Domain
Social Welfare
Civil Rights
Foreign Affairs
X Difference between
Member of Congress
and Partisan
Constituents on
Dyadic Basis
Difference between
All Legislators and
All Districts in the
Aggregate
"Improvement"
of Collective
over Dyadic*
3.26
2.92
4.34
2.45
2.41
2.62
24.8%
17.5%
39.6%
542
Vol. 72
1978
543
likelihood of, say, extreme liberal misrepresentation being canceled out by extreme conservative misrepresentation is greater in a large
assembly than in one of, say, 5 or 10 members.
To illustrate this principle we have computed
the ratios of dyadic to collective representation
for random samples of the districts in the
Miller-Stokes study (Table 5). We should add
that since Miller and Stokes do not collect data
on all Congress members, even the 1.00 sample
is not the population of the House of Representatives; quite likely the full 435 cases as
opposed to the 146 here would show even
greater improvement of representation due to
aggregation.
As expected, the greater the number of
legislators, the more representative the institution as a whole. For the 1 of 10 sample (i.e.,
about 15 cases), dyadic and aggregate representations are nearly identical in social welfare and
civil rights. Aggregate representation shows a
sizable gain in the .5 sample (about 75 districts)
where the figures are quite close to those for
the entire set of districts. Though we cannot
say with any certainty what the dyadic to
aggregate ratio would be for all 435 cases, our
figures here suggest that any improvement
would likely be quite modest (though a very
large assembly would likely give voice to many
more points of view).
Political Control and
Representative Government
Representation, at least as considered in the
context of elections, is usually viewed as a
consequence of political control (i.e., the ability of citizens to remove undesirable officials).16 By means of popular, direct elec-
16Since concepts like "electoral control" are always troublesome,it might be useful to reiteratethat
Issue Domain
Social Welfare
Civil Rights
Foreign Affairs
X Difference between
Member of Congress
and Opposing
Partisan on
Dyadic Basis
Difference between
AlULegislators and
All District Minorities
in the Aggregate
"Improvement"
of Collective
over Dyadic*
3.41
3.21
4.79
2.38
2.20
2.70
30.2%
31.5%
43.6%
544
Vol. 72
1978
545
geneous as possible in their preferences, despite such a relationship is inconsequential. To degeographical constraints. In actual practice this scribe the nature of collective representation is
districting would have to be done not on the not to devalue dyadic representation. Obviousbasis of opinion data, but on more apparent ly, regardless of collective actions, a legislator is
criteria such as economic base, income levels, always electorally accountable to a constituracial and ethnic composition, or other char- ency. This accountability may be imperfect,
acteristics associated with sets of predictable but its existence is not trivial politically. To
political preferences. Under such circumstances, appreciate this, one should consider the posthe odds that each legislator is typical of the sibility of nonelected leaders who could not
district would increase and, if such district be removed regardless of popular objections.
homogeneity were general, the legislature Clearly, such accountability is central to nowould approximate a quota sample of the tions of democratic control of leaders by
population. Needless to say, such criteria could citizens as well as some degree of policy
very well lead to oddly shaped "gerryman- control. Moreover, apart from overall policy
dered" districts that would likely violate cur- representation, purely dyadic representation
rent reapportionment standards established by may very well provide psychological benefits to
citizens who may feel represented by their
the federal courts.
A legislature composed of members chosen particular legislators regardless of the source of
on the basis of quota-sample criteria for dis- representation (or at least enjoy all the attentricting would provide representation on both a tion at election time).
collective and a dyadic basis. That is, while
citizens as a whole are more accurately repreImplications and Speculations
sented by this legislature collectively than they
would be if districts were diverse, it is also true
Collective, as opposed to dyadic, representathat each citizen within the dominant district
tion
is obviously both historically justifiable
group probably receives the most accurate
representation from his or her "own" legislator. and politically possible. We have also suggested
Hence, at least at first glance, the relationship that citizen preferences can indeed be reprebetween the representation of one's opinions sented collectively even if particular legislators
and the voting decision is restored. Neverthe- ignore their constituencies. What we do not
less, it is also likely that where homogeneity of know is whether citizens relate to legislators
opinion prevails, the amount of electoral com- dyadically or as mere components of a colpetition will be very low. It is difficult to lective body. Let us momentarily assume that
imagine a well-developed two-party system and citizens (unlike most political scientists) are less
close elections where a strong consensus exists interested in dyadic legislature relationships
on policy preferences. Even if the legislator in than collective representation. If this were true,
such a homogeneous district did not share the we would make the following predictions about
district's opinion on the salient issues, it would citizen behavior.
First, given that voting for or against 1 of
seem unlikely that this legislator would openly
435
legislators is unlikely to affect greatly the
advocate a contrary view in order to give voters
a real choice.18 In short, if electoral control representation one receives, a lack of citizen
and representation are linked, the control is concern and involvement in legislative elections
likely to be more nominal than threatening, is probably understandable (though citizens do
given safe electoral margins and the lack of an not necessarily engage in the requisite calculus).
After all, it could be argued that it makes little
organized opposition.
sense to get involved in improving one's repreIn making the argument that the dyadic sentation when one cannot even vote on one's
electoral control relationship between citizen best representative and when one's vote, even if
and legislator may not be all that important for absolutely crucial in one's district, affects only
policy representation, we are not claiming that a small fraction of the representative institution. Under such conditions, high involvement
would be worthwhile only if legislative benefits
18Some interesting data on district safeness and could be gained from one's particular legislator
accurate representation are presented in Warren E. (e.g., a private bill). If we assume that people
Miller, "Majority Rule and the Representative System
are interested in what they can affect, we
of Government," in Cleavages, Ideologies and the
also predict that citizen involvement
would
Party Systems, ed. Erik Allardt and Yrjo Littunen
(Helsinki: Academic Bookstore, 1964), 343-76; also would covary with the size of representative
see Morris P. Fiorina, Representatives, Roll Calls, and institutions. Specifically, the fewer the legislators, the greater the impact of each legislator on
Constituencies, pp. 90-100.
546
The AmericanPoliticalScienceReview
Vol. 72
nexus-that citizens can receive benefits without holding much or any electoral power over
legislators. Indeed, the absence of some coercive mechanisms such as a strong two-party
system, close elections, may be associated with
accurate representation. The fact that citizens
still receive benefits (i.e., representation) under
poor control situations, and even when they
can affect only a very small portion of a
legislature's composition, may make citizen
toleration (if not satisfaction) with these
"poor" conditions quite reasonable. This logic
would seem particularly relevant for district
and national minorities who would get "shut
out" completely if legislators slavishly followed
district majorities. No doubt notorious House
iconoclasts like H. R. Gross or Vito Marcantonio represented significant numbers of otherwise unrepresented citizens while probably misrepresenting their district majority.
The benefits of collective representation
might also make more plausible widespread
public indifference towards the idea of responsible party government or other schemes involving greater legislative cohesion.21 The present
system with its complement of mavericks and
party deviates does give expression, no doubt,
to preferences that might very well be excluded
in a system of disciplined, coherent legislative
parties. To a certain extent the existing system
gives citizens the best of both worlds: reasonably accurate representation (even for district
minorities) and legislators who are willing to
perform narrow constituency services. Hence,
to ask citizens to give up the such particularistic
benefits as, for example, legislative intervention
in the bureaucracy, for the promise of better
representation when existing representation is
probably substantial (though not perfect), is to
make an easily refusable offer. Such an offer of
greater party discipline might be more at-
1978
547
tractive if (1) citizens knew the true distributions of preferences in the population; (2) if
popular majorities were systematically violated
in lawmaking by these undisciplined lawmakers,
and (3) the misrepresented citizens believed
that responsible parties, as opposed to "better"
Congress members, would rectify the misrepresentation. Obviously, these requirements are
quite poorly met. In short, given the reasonable
amount of legislative success under present
conditions, a drastic change lacks appeal.