Content uploaded by Joan Barceló
Author content
All content in this area was uploaded by Joan Barceló on Oct 14, 2017
Content may be subject to copyright.
The clarity of the majority’s preference
moderates the influence of lobbying
on representation
Joan Barcelo
´
Washington University in St. Louis, USA
Abstract
Delegate conceptions of representation require activities of legislators to reflect their constituents’ preferences. Recent
research has examined the distortionary effects of lobbying activities on this representational linkage. Here, I argue that
the effect of interest groups on legislators’ behavior depends on the clarity of the majority’s preferences in a district.
When the electorate is narrowly divided, Members of Parliament (MPs) may choose to reap the benefits associated with
interest groups as costs from defection are lowest. The results show that MP defection from constituents’ preferences is
only positively associated with sectional interest group ties when the constituency is narrowly divided on an issue.
Likewise, MP defection is only negatively associated with MP’s ties to cause groups when the constituency is narrowly
divided on an issue. These results are important because they specify the conditions under which interest group lobbying
is sufficient to override constituents’ preferences.
Keywords
interest groups, lobbying, representation
A crucial aspect in the chain of democratic responsiveness
is that the representatives’ activity should reflect their con-
stituents’ preferences (Pitkin, 1967; Powell, 2004, among
others). Although an idealized vision of democracy might
suggest that the link between voters and representatives
should be straightforward, there are strong reasons to
believe that this is not always the case. In two recent arti-
cles of the greatest theoretical importance, Portmann et al.
(2012) and Giger and Klu
¨ver (2016) provide a rather pes-
simistic view of the linkage between voters and represen-
tatives. Both suggest that the connection is distorted by the
lobbying activities of interest groups.
Both groups of authors empirically show that interest
groups indeed appear to exert strong influence on legisla-
tors’ behavior in Parliament. First, Portmann et al. (2012)
(henceforth, PSE) showed that those Members of Parlia-
ment (MPs) who have more ties with interest groups are
more likely to vote against their constituents. Building
upon this work, Giger and Klu
¨ver (2016) (henceforth,
GK) divide interest groups depending on the nature of their
aims and report that those MPs who have strong ties with
“sectional” groups, groups that represent specific segments
of a society and pursue private interests, such as farmers or
business organizations are more likely to defect from the
preferences of their voters. By contrast, legislators who
have strong ties with “cause” groups, those that pursue a
common good such as human rights or a clean environ-
ment, are more likely to vote with the preference of the
majority of their constituents. This finding regarding sec-
tional group influence is troubling for the notion of repre-
sentational democracy in the presence of lobbying, and it
becomes even more disturbing in the light of authors’ data
showing the frequent disagreement between voters and
interest groups on most referendums in Switzerland
throughout the period of study: 1996, 2006, and 2009.
1
This research complements these striking findings by
bringing the clarity of citizens’ preferences into the analy-
sis.
2
I argue that the clarity of the majority’s preferences, or
constituents’ ideological heterogeneity, is a mechanism to
balance the (often) competing interests between voters and
Paper submitted 24 March 2017; accepted for publication 26 May 2017
Corresponding author:
Joan Barcelo
´, Campus Box 1063, Department of Political Science,
Washington University in St. Louis,OneBrookingsDrive,St.Louis,
MO, 63130-4899.
Email: joanbarcelosoler@wustl.edu
Party Politics
1–9
ªThe Author(s) 2017
Reprints and permission:
sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/1354068817715803
journals.sagepub.com/home/ppq
interest groups. Specifically, MPs respond to both voters
and interest groups, but the influence of interest groups is
conditional on the divisiveness of the electorate on a policy
issue. When citizens are narrowly divided on an issue, the
electoral impact from defection—the difference between
the number of voters who will reward or punish the legis-
lator for her behavior—is smaller, which allows interest
groups to exert a greater influence on legislators’ behavior.
By contrast, when citizens are clearly favoring one side of a
policy, legislators’ reward (or punishment) as a result of
following the majority (minority) is larger, which decreases
the influence of interest groups on legislators’ behavior. It
is also possible that when the electorate in a district is
narrowly divided on an issue, MPs are more likely to err
when estimating the preferences of the median voter among
their constituents.
3
Using a data set that includes
individual-level results of 20,260 legislative votes cast in
the Swiss Parliament by 488 MPs on 118 policy proposals
subject to parliamentary approval in the period 1996–2009
and referendum outcome, I find empirical evidence for
these propositions.
This research contributes to a long line of political
science research that examines the relationship between
heterogeneous constituencies and elite-mass congruence
by suggesting that the opportunities for rent extraction
might be an explanation for the widespread finding that
legislators from more homogeneous states are more
responsive to their constituents (e.g., Bishin et al., 2006;
Gerber and Lewis, 2004; Gulati, 2004). Hence, constitu-
ents’ heterogeneity in policy preferences crucially affects
the capacity of interest groups to influence the behavior of
legislators and, thus, affecting the congruence between
legislators and their constituents.
In what follows, I first illustrate the mechanics of the
argument by developing a simple decision model of any
given legislator. The model elucidates the differences
between the currently dominant theory on the influence
of lobbying on legislative behavior, and the impact of
lobbying on representation that considers the clarity of
citizens’ demands. Second, I discuss the data and meth-
ods most appropriate for testing the distinct empirical
implications from both theories. Third, I subject those
data to an empirical examination. Finally, I conclude the
research note by arguing that breakdowns in the repre-
sentation of constituents’ interests occurs in only very
limited circumstances.
Two theories of representation
in the presence of lobbying
Much of the recent literature trying to empirically gauge
the impact of lobbying on representation has ignored the
strength of the clarity of constituents’ preferences. In the
theoretical framework of both PSE and, especially, GK,
MPs associated with sectional (cause) interest groups are
always more (less) likely to defect from their voters. In
other words, according to GK, the ties with interest groups
influence defection irrespective of the number of citizens
on each side of the policy question. To clarify the basic
components, the following utility equation illustrates their
theory of representation in the presence of lobbying: the
constant-effect hypothesis (PSE, 2012; GK, 2016).
4
Let lbe a legislator who decides whether to defect from
his or her citizens or to represent them by voting in favor of
their interests. A legislator u
l
maximizes the following util-
ity equation:
ulðdefectionÞ¼BþIs;ð1Þ
ulðcongruenceÞ¼ BþIc;ð2Þ
where Brefers to the size of the electoral impact that the
roll-call vote will have on legislator l, which is non-
negative: B0; I
s
denotes the sum of contributions that
sectional interest groups provide to legislator lin exchange
of defecting from his or her constituents, which is non-
negative I
s
0; and I
c
denotes the sum of contributions
that cause interest groups provide to legislator lin exchange
of not defecting from his or her constituents, which is non-
negative I
c
0.
5
In words (a) the utility of defecting is
equivalent to the rents extracted from the interest groups
minus the electoral punishment and (b) the utility from
congruence is the electoral reward from congruence plus
the benefit obtained from supporting the policy that favors
cause interest groups. Consequently, a legislator lwill
defect only if: I
s
>2BþI
c
.
In the light of the above utility function, PSE’s and GK’s
theoretical models implicitly assume that legislators’ deci-
sions to defect entirely depend on B,I
s
, and I
c
. On the one
hand, as the size of the expected contribution from sec-
tional interest groups due to defection, I
s
, increases, defec-
tion becomes more likely. Similarly, as the size of the
expected contribution from cause groups due to congru-
ence, I
c
, increases, defection becomes less likely. GK use
the number of each type of interest groups associated with
the legislator to operationalize the size of the expected
contribution from interest groups.
On the other hand, the likelihood of defection increases
as the electoral reward/punishment, B, from congruence/
defection increases. Yet, the existence and relevance of B
remains underdeveloped in the dominant model of repre-
sentation in the presence of lobbying. The above theoretical
approach assumes that once a legislator has voted with the
expected median voter—the expected majority’s prefer-
ence—the legislator obtains a fixed reward B. This consti-
tutes the constant-effect hypothesis:
H1 (constant effect): AnMPstieswithsectional
(cause) interest groups is positively (negatively) associ-
ated with defection irrespective of the size of the
majority.
2Party Politics XX(X)
An alternative perspective would suggest that voters do
not reward or punish legislators with respect to their con-
gruence with the entire district but with respect to their
own interests. If we understand that voters electorally
reward legislators who favor their interests but punish
legislators who do not favor their interests, then the clarity
of the majority’s preferences should be a crucial determi-
nant in legislators’ defection. Clearly, this should condi-
tion the influence of interest groups. A more complete
theory of representation in the presence of lobbying
should consider the size of the majority group because its
size is proportional to the magnitude of both the electoral
reward from congruence and the electoral punishment
from defecting.
To consider this more completely, I bring in the clarity
of citizens’ preferences on a policy issue into the analysis.
This sheds light on the fundamental role of voters in deter-
mining when the role of interest groups is influential in the
legislators’ decisions. When citizens are narrowly divided
on an issue, the electoral reward or punishment from con-
gruence or defection is smaller, which allows interest
groups to exert a greater influence on legislators’ behavior.
By contrast, when citizens clearly favor a policy outcome
over another, a legislators’ reward (or punishment) as a
result of following the majority (minority) is larger,
decreasing the influence of interest groups.
To clarify the argument, I expand the above utility func-
tion by reflecting this alternative theory of representation in
the presence of lobbying. Thus, let lbe a legislator who
decides whether to defect from his or her constituents or to
represent them by voting in favor of their interests. In the
decreasing-effect hypothesis, a legislator lmaximizes the
following utility equation:
ulðdefectionÞ¼BEðcÞþIs;ð3Þ
ulðcongruenceÞ¼BEðcÞþIc;ð4Þ
where E(c)¼E(p
w
(1 p
w
)), p
w
denotes the relative size
of the majority, which ranges between 0 and 1: p
w
2(0,1];
and, thus, E(c) indicates the expected difference between
the majority and the minority—the clarity of the majority’s
preference—which ranges between 0 and 1: c2(0,1]. The
solution to the new utility function is that legislator lwill
defect only if I
s
>2Bc þI
c
. Otherwise, she will vote con-
gruently with her constituents’ preferences.
The theoretical contribution of this new setting lies on
the conditional nature of B.Bc captures the idea that a
legislator’s electoral rewards (if negative, electoral punish-
ment) for voting in favor of (if negative, against) the major-
ity of the citizens in his or her district is expected to be
proportional to the share of voters who voted the same as
the legislator l. Similarly, as E(c), the expected relative size
of the majority in a district, becomes smaller, legislator l
becomes more uncertain whether she will vote congruently
in each policy with the median voter of her district or not.
In the extreme case in which the electorate is totally
divided, E(c)¼0, the solution to the inequality becomes
I
s
>I
c
. In this situation, lobbying activities entirely dom-
inate MPs’ legislative behavior.
Hence, the electoral reward/punishment from defection
becomes larger as the clarity of the majority’s preference,
and so the clarity of the signal about what the citizens want,
increases as legislators should be less influenced by the
preferences of interest groups. In the case of sectional
groups, the positive association between interest group ties
with defection is expected to become weaker as the major-
ity’s preference becomes clarified—the impact of I
c
on
defection decreases as E(c) increases because this means
that the costs of punishment from voting against MPs con-
stituents increases.
In the case of cause groups, the expected effect is oppo-
site because a clear majority’s preference simply adds
greater incentives for not defecting from the electorate.
Thus, the role of the clarity of the constituents’ preferences
is expected to reinforce the effect of cause groups. There-
fore, the clarity of the majority’s preference should make
legislators even less likely to defect from their constituents.
The two forces—the effect of the majority’s preference and
cause groups—move in the same direction and, thus, we
should expect reinforcement effects. Yet, their reinforcing
effects should be weaker because the reinforcement process
may be empirically unobserved in many situations because
they may be empirically equivalent. In short, the reward
from voting in favor of many constituents or cause groups
may be sufficient to discourage defection if either the
majority clearly favors one side of the policy issue or the
contributions from cause groups are sufficiently large. This
leads to an alternative hypothesis:
H2 (decreasing effect): the association of MPs ties with
interest groups with defection becomes stronger as the
clarity of the majority’s preference decreases.
Research design
This section lays out how the hypotheses are tested empiri-
cally. I first describe some key institutional features of the
Swiss political context and the dataset that allows me to test
the theories. Then, I discuss the measures of the dependent,
independent and control variables. Finally, I discuss the
estimation strategy.
The Swiss context
The Swiss context provides an ideal setting to test theories
of representation. The Federal Assembly is bicameral,
being composed of the 200-seat National Council—the
lower chamber—and the 46-seat CouncilofStates—the
upper chamber. The members of the National Council
(Nationalrat or Conseil National)areelectedfor4years
Barcelo
´3
in a system of proportional representation. The 26 elec-
toral districts correspond to the 26 cantons of the federa-
tion, and their magnitude varies as a function of the
number of voters in each district, with a minimum of one
seat per district. The largest districts are Zurich and Bern
with 35 and 25 seats, respectively, and 6 districts have a
single member.
Just like most democracies around the world, Swiss MPs
vote for the creation and modification of laws and consti-
tutional amendments. When making their roll-call votes in
parliament, MPs do not have exact knowledge of constitu-
ent preferences. The unique feature of the Swiss system is
the importance of the popular vote. Switzerland combines
their representative democracy with features of direct
democracy by which legislative activity is directly subject
to approval by citizens via three types of popular votes:
mandatory referendums, constitutional amendments or
adhesion to supranational communities require a referen-
dum; facultative referendums, laws are subject to referen-
dum if 50,000 citizens demand it; and, finally, popular
initiatives, 100,000 citizens may demand for a constitu-
tional change, which will be discussed in the Parliament
and voted in a referendum.
6
Due to the Swiss institutional configuration, referen-
dums regularly reveal the preferences of constituents,
which can then be used to match the revealed preferences
of constituents with what the MPs voted in the Parliament
on the exact sane bill. The simultaneous use of parliamen-
tarian procedures and direct democracy over identical pol-
icies provides a setting for researchers to locate MPs’ and
citizens’ preferences on the same scale. Along with prior
studies in Switzerland and elsewhere, I exploit the match
between the roll-call votes and the popular votes to directly
associate the revealed preferences of the constituents with
the actions of their representatives.
Data
In order to test the hypotheses, I use the legislative
voting and interest groups data from the Swiss Lower
House used by GK (2016). This data set builds on the
work by Portmann et al. (2012) and Stadelmann et al.
(2013, 2015), which has become one of the most pro-
minent data contributions to the literature of represen-
tation in the last years. Specifically, the data set
includes individual-level results of 20,260 legislative
votes cast in the Swiss Parliament by 488 MPs on
118 policy proposals subject to parliamentary approval
in the period 1996–2009. To assess the defection of
MPs, the data set combines from roll-call votes in the
Parliament with the results of the popular vote in each
MPs’ electoral district on the same policy issues. The
advantage of the setting is that, as in countries without
referendums, Swiss politicians do not have exact knowl-
edge of constituent preferences when making decisions
in parliament, which makes the results particularly cred-
ible and generalizable.
Measures
For the empirical analyses, the dependent variable is defec-
tion, an indicator that equals one when a member of a
parliament does not vote the same way as the majority of
his or her district. The main independent variables are the
number of sectional groups and the number of cause groups
associated with the MP. A unique institutional feature of
the Swiss system is that all MPs are required to report all
interest groups to which they belong to, which are then
published in an annual interest groups registry.
7
These vari-
ables capture the number of interest groups that MPs indi-
cated in the parliamentarian registry in the year of the
referendum.
8
The conditional variable of interest, E(c), is
captured by the ultimate margin of victory of the referen-
dum. The margin of victory is measured as the percentage
vote of the winning outcome 50.
To clarify the distinct assumption made in prior work
using this data set and the assumption made here, it should
be noticed that a crucial aspect in the data set is the
sequence of events in the revelation of preferences between
legislators and citizens. The assessment of MP defection is
based on the roll-call votes by the MPs in the Parliament on
the policy, which takes place prior to the referendum where
the constituents’ preferences are revealed. If the MP does
not vote the same as the median voter subsequently votes in
her district, this is coded as defection. Yet, legislators cast
their votes before the revelation of constituents’ prefer-
ences through the referendum ever takes place.
Thus, to interpret any disjuncture between constituents
and MPs as “defection,” the analyses based on the constant-
effect hypothesis have assumed that MPs can perfectly pre-
dict the outcome of the referendums with certainty. In fact,
PSE and GK are aware of this and explicitly argue that MPs
are quite knowledgeable of the opinion of their constituents
through public debates and opinion polls. The analysis of
the decreasing-effect hypothesis relaxes this assumption.
Specifically, the decreasing theory admits that all that an
MP may know about a referendum that is ultimately
decided by a narrow (lopsided) margin of victory is that
her voters are (not) divided on that issue. Note that the task
of forecasting the outcomes of referendums is particularly
daunting given that the outcome, the constituency referen-
dum outcome, is measured at the level of the electoral
district (the canton), where the informational environment
is probably rather sparse. Therefore, the assumption made
by the decreasing hypothesis is more realistic, considering
what legislators may not know about a closely contested
referendum before it ever takes place.
As with previous studies, the models include controls for
the number of months until next election, the number of
MPs per canton, whether the MP votes with his or her party,
4Party Politics XX(X)
the margin of victory in the parliamentarian roll-call vote,
the salience of the vote,
9
and two dummies that indicate
the type of the referendum, whether it was obligatory,
facultative, or the result of a popular initiative (category
of reference).
Estimation strategy
The aim of the empirical section is to measure the effect of
sectional and cause interest groups on defection at different
values of the clarity of the majority’s preferences. To do so,
I begin by estimating a mixed effects logistic regression
with random intercepts by party and canton in which the
margin of victory in the canton of the MP is interacted with
the number of sectional interest groups at the MP level.
Although the mixed effects model provides efficient
estimates, their consistency hinged on the exogeneity
assumption; that is, the assumption that the residuals are
independent of the covariates. Because this might not hold,
I further test the robustness of the main specification by
adding referendum, party, and canton fixed effects. All
variance at the level of the referendum, party, and canton
are controlled out with this set of fixed effects, and, con-
sequently, it minimizes the potential for omitted variable
bias. In addition, I compute cluster-robust standard errors at
the level of the MP to correct the correlation of observa-
tions for the same MP across different referendum.
Results
To test whether margin of victory of the referendum mod-
erates the effect of interest groups on defection, I show the
Table 1. The effect of lobbying on MP defection conditional on the margin of victory of the referendums.
Dependent variable: defection
Model 1 Model 2 Model 3 Model 4 Model 5
Number of sectional groups 0.013
**
0.086
***
0.085
***
0.093
***
0.093
***
(0.005) (0.009) (0.009) (0.009) (0.015)
Number of cause groups 0.035
***
0.045
***
0.043
***
0.044
***
0.044
***
(0.008) (0.015) (0.015) (0.016) (0.024)
Margin of victory 0.034
***
0.023
***
0.023
***
0.074
***
0.074
***
(0.002) (0.002) (0.002) (0.013) (0.013)
Number of sectional groups margin of victory 0.006
***
0.006
***
0.007
***
0.007
***
(0.001) (0.001) (0.001) (0.001)
Number of cause groups margin of victory 0.001 0.001 0.001 0.001
(0.001) (0.001) (0.001) (0.002)
MP-party-level controls
Official party position congruence 0.551
***
0.546
***
0.568
***
0.684
***
0.684
***
(0.049) (0.049) (0.050) (0.059) (0.087)
Referendum-level controls
Closeness of parliament decision 0.043
***
0.044
***
0.044
***
(0.002) (0.002) (0.002)
Salience 0.003
*
0.003
*
0.003
**
(0.002) (0.002) (0.002)
Number of months until next election 0.003
***
0.003
***
0.003
***
(0.001) (0.001) (0.001)
Referendum obligatory 0.423
***
0.436
***
0.436
***
(0.049) (0.049) (0.049)
Referendum facultative 0.511
***
0.512
***
0.513
***
(0.040) (0.040) (0.040)
Canton-level controls
Number of MPs in district 0.005 0.005
(0.004) (0.004)
Party RE/FE RE RE FE FE FE
Canton FE/RE RE RE FE FE FE
Referendum FE/RE N N N FE FE
Observations 20,260 20,260 20,260 20,260 20,260
Log likelihood 11,593 11,541 11,480 10,729 10,729
AIC 23,212 23,113 23,074 21,789 21,789
Note. Constants are omitted from the output. Models 1—4 report standard errors in parenthesis and Model 5 adjusts the standard errors and reports
robust-clustered standard errors at the level of the MP in parenthesis. Referendum-level controls and canton-level controls are excluded models that
include referendum and canton fixed effects, respectively.
*
p< 0.1;
**
p< 0.05;
***
p< 0.01.
Barcelo
´5
main-effect model, but I then interact the main effects with
the margin of victory of the referendum. Model 1 in Table 1
presents the main effects of interest group ties. It shows the
effect of the number of sectional and cause groups associ-
ated with the MP on defecting from their electorate. Model
2 complements the main effects by adding an interaction
term of the number of interest groups with margin of vic-
tory of the referendum—or the clarity of the signal from the
constituents on what position they want the MP to take on
the issue.
The first element to consider from the results is that the
added interaction in the models significantly improves their
goodness of fit (LR test ¼w
2
(2) ¼102.8, pvalue < 0.001).
This can be seen in the results of the likelihood ratio test,
which shows that the addition of the interactive terms sig-
nificantly improves the fit of the models over the null
model treating the effects of interest groups on defection
as constant across all referendums margins of victory.
From the regression results in Table 1, there is suffi-
cient evidence to reject hypothesis 1, the constant-effect
hypothesis, which conjectured a constant effect of ties
with interest groups on defection along the margin of
victory of the referendums. Instead, sectional group ties
increase the probability that MPs defect from their voters
but only when the electorate is narrowly divided. The
main effect of the number of sectional groups from mod-
els 2–4 indicate that when the margin of victory is 0, in the
extreme case when the referendum outcome is ultimately
50/50, then the effect of lobbying on the MP voting
against their constituents is statistically significant in the
expected direction at a 99%confidence level. In addition,
we also see that the interaction term takes a positive value.
This indicates that as referendums become more lopsided,
the effect of lobbying dissipates.
Consistent with hypothesis 2, the decreasing-effect
hypothesis, MPs are more likely to “defect” from a majority
of voters when the electorate is narrowly divided or the
outcome of the subsequent referendums may appear
uncertain. Either MPs misjudge where the median voter
on the topic stands or they reason that their roll-call vote
will only alienate an acceptably small number of voters. By
contrast, MPs become less likely to defect from their voters
as the referendum becomes more lopsided or, in other
words, when the clarity of the signal on what the electorate
wants becomes more unequivocal toward a specific posi-
tion on the issue. In these instances, MPs are less likely to
err in estimating the district median voter and she is unwill-
ing to risk alienating such a large group.
To illustrate the conditional effects of interest group ties,
Figure 1 reports the marginal effects of the number of sec-
tional groups tied to an MP on the MP’s defection from his
or her electorate at different values of margin of victory of
the referendum. Panel (a) in Figure 1 shows that the most
striking finding from GK’s original research piece, the pos-
itive association between MP’s ties with sectional groups
and MP’s defection, is statistically significant in the
expected direction for referendums with a margin of vic-
tory of 63.1%/36.9%or less. Referendums with margins of
victory this narrow constitute 37.9%of the observations
(those observations with an estimated effect above the
dashed line in Figure 1, Panel (a)). Even for the referen-
dums for which the association is positive, the effect of
lobbying on defection consistently decreases with the lop-
sidedness of the referendums. Hence, the main finding rap-
idly becomes substantively insignificant or lose its
statistical reliability once the district’s preferences become
clear (or the referendum becomes mildly lopsided). In other
words, ties to sectional groups increase the likelihood of
voting with the district’s majority.
Likewise, MPs ties to cause group strengthen the link
between MPs and voters, but the strengthening effect
declines as the referendum becomes more lopsided. Model
2 shows that the coefficient of the main effect of the num-
ber of cause groups on defection is negative. This indicates
that as the number of cause groups associated with an MP
increases, MPs are less likely to defect from the electorate
Figure 1. Marginal effect of interest groups on defection by referendum margin of victory. (a) sectional groups and (b) cause groups.
6Party Politics XX(X)
when the margin of victory is very narrow. As the electo-
rate becomes more unequivocally in favor of one side of the
issue, the referendum becomes more lopsided, the effect of
cause interest groups vanishes. Panel (b) in Figure 1 shows
the decline of the effect of cause groups on defection. More
precisely, the negative association between MP’s ties with
cause groups, and MP’s defection is statistically significant
in this direction for referendums with a margin of victory of
74.2%/25.8%or lower. Although these referendums con-
stitute most of the observations in the sample—about 78%
of the observations—the effect of cause-group lobbying on
defection strikingly weakens as a referendum ultimately
ends up lopsided—or as the signal of the majority’s pre-
ference is clarified.
In sum, there is evidence to reject hypothesis 1, which
states that the effect of lobbying by cause groups on MPs’
defection is constant along the values of the margins of
victory of the referendums. Instead, there is evidence that
cause groups’ lobbying strengthens the MP–citizen link,
lowering the chances of defection, only when the electorate
is narrowly divided, yet their effect substantively weakens
when the signal of the majority’s preference of the electo-
rate becomes more obvious.
Robustness checks
A potential concern from these model specifications is that
there might be some unobserved heterogeneity across par-
ties, cantons, and referendums that might explain these
above results. First, political parties constitute another
important principal for MPs beyond voters and interest
groups. Thus, some MPs may be affiliated with some inter-
est groups because their party membership requires them to
be affiliated with certain interest groups or they may seek
congruence with their party instead. A second concern is
that some cantons, such as larger cantons, might be more
central to the political system and, therefore, more likely to
have ties with interest groups and distinct rates of defec-
tion. Finally, a third type of argument could be made that
the characteristics of referendums might affect the lopsid-
edness of a referendum, the strength of the interest groups’
influence as well as the likelihood of defection.
To alleviate these concerns, I proceed in two ways. First,
models include an MP-level control for relevant observed
confounders, namely, MPs’ congruence with the official
position of the party, and a set of referendum-level control
variables. The inclusion of observed controls should reduce
the concerns with respect to underlying omitted variable
bias, although it falls short of entirely removing that con-
cern. Second, model 3 includes party and canton fixed
effects, and models 4 and 5 further add referendum fixed
effects. These model specifications control for unobserved
omitted variables that are invariant across MPs that belong
to the same party, unobserved omitted variable that are
invariant across MPs from the same canton, and
unobserved omitted variables that are invariant across
observations in a referendum. In addition, model 5 clusters
the standard errors at the MP level to correct for the fact
that behavioral observations of the MPs over time across
different referendums are not independent. In other words,
these models provide the most stringent test to my hypoth-
esis because they focus on within-party, within-canton, and
within-referendum variation alone and, thus, alleviate all
concerns of omitted variable bias with respect to party-,
canton-, or referendum-specific characteristics.
In the main analyses, I follow the most recent contribu-
tion in this area by GK (2016) to generate the total number
of cause and sectional groups. Because the theoretical sec-
tion conjectures that the effects of each type of groups will
go in distinct directions, using the total number of interest
groups would mean that the effects of distinct types of
groups would cancel each other out. In practice, however,
cause and sectional interest groups might not always be
easy to classify. As a robustness check, Table 1A reana-
lyzes the main models but using the total number of interest
groups. As expected, the main coefficients and the interac-
tion effect are generally smaller, but all remain in the
expected direction and statistically significant. This
evidence provides further empirical support for the
decreasing-effect hypothesis.
A remaining concern is that the official and legal task of
politicians in the National Council is to represent the entire
nation although they are elected within their constituency.
Consequently, Table 2A checks the robustness of the main
findings with the outcome variable being defection with
respect to the entire nation. Given the high correlation
between the two measures,
10
it is not surprising that
although the magnitude is generally smaller, the results
remain substantively unaltered.
Finally, one might consider that the decreasing-effect
hypothesis is particularly relevant among those districts
in which the district magnitude is small, as the connection
between the representatives and the median voter is stron-
ger (Portmann et al., 2012). Hence, we should be concerned
about the mechanisms underlying the main findings pre-
sented above if we find out that the evidence exclusively
hinges on electoral districts with large magnitudes. To test
for this, Table 3A in the online Appendix reestimates the
main models only for the sub-sample of cantons with five
seats or less.
11
Results remain substantively unaltered for
representatives of this subgroup of districts.
Conclusion
Whether and when MPs defect from their constituents to
support the preferred positions of certain types of interest
groups is more nuanced than previously portrayed. The
analysis offered in this article extends prior work on the
influence of lobbying on representation. Specifically, I
argued that a divided electorate is the context that
Barcelo
´7
minimizes the costs of defection, allowing MPs to defect
from their constituents to reap the benefits of pleasing the
interest groups’ demands. Using widely used data from
Swiss roll-call votes matched with popular vote, I show
that MPs respond to both voters and interest groups, but
the people’s voice dominates interest groups’ demands
when the electorate in a district is clearly on one side of
a policy issue. By contrast, interest groups are more likely
to shape policy decisions when the electorate is narrowly
divided.
These findings have important implications for theories
of democratic representation, since they specify under
which circumstances interest group lobbying influences
legislative behavior. Specifically, they contribute to the
literature on congruence and representation by showing
that a narrowly divided electorate facilitates the distortion-
ary effects of lobbying on representation. Citizens who live
in districts in which there is usually a clear majority’s pre-
ference for one side of policy issues are more likely to have
their interests represented than similar citizens who live in
districts in which there is much disagreement over policy
positions. From the point of view of interest groups, they
are more likely to influence policy and have their prefer-
ences represented by MPs—even at the expense of voters’
preferences—in those districts in which citizens have very
heterogeneous policy positions. In contrast, interest groups
are less likely to influence MPs behavior in districts areas
where citizens have a clear position over the issue. Conse-
quently, the relative dispersion of interests in a district
helps characterize the extent to which lobbying may distort
the chain of democratic responsiveness between voters and
representative, as well as the effectiveness of lobbying on
shaping MPs’ behavior.
Furthermore, this article offers some avenues for future
research. Further work may explore additional conditions
under which interest groups are more likely to influence the
responsiveness of legislators to their constituents such as
the technical matters of the issues or their timing in the
electoral cycle. Just as most work in the congruence liter-
ature,
12
I measure ideological congruence and the clarity of
the majority’s preferences at the level of cantonal consti-
tuencies. Yet, it is also possible that defection and the
clarity at the level of the MPs’ own partisan constituency
is of the highest relevance. Therefore, further work might
extend this work by using finer grained data with observed
preferences at the level of partisan cantonal constituencies.
In addition, the lack of direct measure of the preferences
of interest groups in the referendums is a limitation of this
work that also constitutes an opportunity for future
research. Following Giger and Kluver (2016), this article
builds on the assumption that the preferences of sectional
(cause) groups are, on average, more likely to be contrary
(in favor) of the preferences of the constituents. While this
might be true on average, measuring the preferences of
interest groups in every referendum might be a move
forward to ensure that MPs defection from citizens is
indeed geared to please the interest groups. Finally, exter-
nal validity of these results beyond the case of Switzerland
is difficult to assess from a single-country study. Conse-
quently, I hope that further research is carried out in com-
parable countries to assess the extent to which the clarity of
the majority’s preferences conditions the role of lobbying
on legislators’ congruence beyond the case of Switzerland.
Overall, this study makes an important contribution as it
specifies a condition under which interest group lobbying
suffices to override constituents’ preferences, but it also
opens more avenues for future research when, why, and
how lobbying affects the chain of democratic responsive-
ness in democratic political systems.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with
respect to the research, authorship, and/or publication of this
article.
Funding
The author(s) received no financial support for the research,
authorship, and/or publication of this article.
Supplementary material
The online appendices are available at: http://ppq.sagepub.com/
content/by/supplemental-data.
Notes
1. These results are based on proportionally elected politicians,
from the Swiss Lower House of Parliament, see Stadelmann
et al. (2016) for a replication paper using the majority-elected
politicians from the Upper House. Findings do not extend to
the Upper Chamber.
2. For an argument on the size of the winning majority and the
extraction of political rents in another setting, see Kauder and
Portrafke (2016).
3. This argument is consistent with the evidence showing that
parties and candidates are less able to identify the policy
platforms that voters prefer in heterogeneous constituencies
(see Bailey and Brady, 1998; Ensley et al., 2009).
4. The current model depicts MPs as the agents of two princi-
pals: the interest group and their constituents. However, other
stakeholders in the process are political parties. Even though
the specification of the relationship between MPs and their
parties is beyond the scope of this paper, I do account for
them in the empirical section by adding a party-congruence
control and party random and fixed effects (see the Research
Design section for further details).
5. This simple utility equation assumes that the preferences of
sectional interest groups are opposed to the interests of the
expected majority of citizens and cause groups support the
interests of the expected majority of citizens.
6. See Kriesi (2012) for further details on the Swiss direct-
democracy institutions and procedures.
8Party Politics XX(X)
7. Among those MPs who indicated a tie with an interest group,
96%reported to hold an executive function (e.g., board
member or president).
8. See GK (2016: 195–197) for further details on the construc-
tion of these variables.
9. The measure of salience was originally constructed by the
media attention to the proposal by counting the number of
mentions in leading newspaper Neue Zurcher Zeitung articles
in the 3-month period before and after the referendum, see
GK (2016) for more details.
10. The correlation between the two measures of defection in the
dataset is 0.78.
11. They constitute the 14.4%of the observations in the sample.
12. See Golder and Stramski (2010) for a review of congruence
measures.
References
Bailey M and Brady DW (1998) Heterogeneity and representa-
tion: the senate and free trade. American Journal of Political
Science 42(2): 524–544.
Bishin BG, Dow JK and Adams J (2006) Does democracy
“suffer” from diversity? Issue representation and diversity in
senate elections. Public Choice 129(1): 201–215.
Ensley MJ, Tofias MW and De Marchi S (2009) District complex-
ity as an advantage in congressional elections. American
Journal of Political Science 53(4): 990–1005.
Gerber ER and Lewis JB (2004) Beyond the median: voter pre-
ferences, district heterogeneity, and political representation.
Journal of Political Economy 112(6): 1364–1383.
Giger N and Klu
¨ver H (2016) Voting against your constituents?
How lobbying affects representation. American Journal of
Political Science 60(1): 190–205.
Golder M and Stramski J (2010) Ideological congruence and elec-
toral institutions. American Journal of Political Science 54(1):
90–106.
Gulati GJ (2004) Revisiting the link between electoral competi-
tion and policy extremism in the US Congress. American
Politics Research 32(5): 495–520.
Kauder B and Potrafke N (2016) Supermajorities and political
rent extraction. Kyklos 69(1): 65–81.
Kriesi H (2012) Direct democracy: the Swiss experience. In:
Brigitte G and Kenneth N (eds) Evaluating Democratic
Innovations: Curing the Democratic Malaise.NewYork,
NY: Routledge, pp. 39–55.
Pitkin HF (1967) The Concept of Representation. Berkeley, CA:
University of California Press.
Portmann M, Stadelmann D and Eichenberger R (2012) District
magnitude and representation of the majority’s preferences:
evidence from popular and parliamentary votes. Public Choice
151(3-4): 585–610.
Powell GB (2004) Political representation in comparative politics.
Annual Review of Political Science 7(1): 273–296.
Stadelmann D, Portmann M and Eichenberger R (2013) Quantify-
ing parliamentary representation of constituents’ preferences
with quasi-experimental data. Journal of Comparative Eco-
nomics 41(1): 170–180.
Stadelmann D, Portmann M and Eichenberger R (2015) Income
and policy choices: evidence from parliamentary decisions and
referendums. Economics Letters 135: 117–120.
Stadelmann D, Portmann M and Eichenberger R (2016) How
lobbying affects representation: results for majority-elected
politicians. The BE Journal of Economic Analysis & Policy
16(4): 1–8.
Author biography
Joan Barcelo
´is a PhD student at Washington University in
St. Louis. He is interested in the study of representation, public
opinion, and comparative political behavior. He has authored or
co-authored some 10 articles in peer-reviewed academic journals
and 3 chapters in academic books.
Barcelo
´9
