Strategic behavior in parliament ∗

Sarah Butikofer†and Simon Hug‡

CIS - ETH ZurichDepartement de science politique, Universite de Geneve

First version: August 2008, this version: May 13, 2010

Abstract

Most research on roll call votes considers each voting decision by mem-bers of parliaments (MPs) as an independent observation. Only recentlyhave scholars (e.g., Clinton and Meirowitz, 2004; Clinton, 2007) startedto assess how knowledge about the sequence of votes may help us to un-derstand the legislative process more in detail. Many of these analysesare, however, predicated on quite important assumptions regarding theforward-looking capacities of MPs.

In this paper we draw on this more recent literature and bring it to bearin an analysis of two bills adopted in the Swiss parliament. Having availabledetailed information on the MPs preferences over various options voted onwe are able to test whether MPs behave strategically and to what degreethey are capable of anticipating the way forward through the agenda tree.We find evidence that MPs behave strategically, however, their foresight isnot as perfect as we would expect from theoretical models.

∗ An earlier version of this paper was presented at the Annual Meeting of the AmericanPolitical Science Association in Boston (August 28-31, 2008), a seminar at the University ofMannheim and at the Annual Meeting of the Midwest Political Science Association in Chicago(April 2-5, 2009). Thanks are due to the participants at these events, especially Kenneth Shepsleand David Hugh-Jones, for their helpful comments, as well as to the Swiss Parlamentsdiensteand especially Ernst Firschknecht and Andreas Sidler, who provided the main bulk of theempirical data used here and helped us using it. Tobias Schulz provided excellent researchassistance, Stefanie Bailer, Michael Cemerin, Flavia Fossati, were an invaluable help in carryingout our MP-survey and Stefanie Bailer also provided helpful comments. The Swiss NationalScience Foundation (Grant No 100012-111909) provided generous funding.† Global Governance; CIS/ ETH Zurich; Seilergraben 49; 8001 Zurich; Switzerland; email:

sarah.buetikofer@gg.gess.ethz.ch‡ Departement de science politique, Faculte des sciences economiques et sociales; Universite

de Geneve; 40 Bd du Pont d’Arve; 1211 Geneve 4; Switzerland; phone ++41 22 379 83 78;email: simon.hug@politic.unige.ch

1

1 Introduction

Research on roll call votes has made considerable headway over the last decade or

so. Sophisticated methods to estimate ideal-points on the basis of roll call data

have been extensively developed (for a review see Poole, 2005). Similarly work

on the effect of parties and accountability (see for instance Cox and McCubbins,

2005; Sieberer, 2006; Carey, 2008) makes innovative use of the roll call vote record.

Most of this research is based, however, on the assumption that individual roll call

votes are independent of each other. Implied in this assumption is that members

of parliament (MPs) vote sincerely and not in a sophisticated manner.1

While this latter assumption might be defended on the basis of Krehbiel and

Rivers’s (1990) and Groseclose and Milyo’s (2009, 2010) critique of work on

sophisticated voting in the US Congress, this critique is based on elements that are

quite specific to this latter institution. Similarly, more recent work on strategic

voting (e.g., Clinton and Meirowitz, 2004; Leemann, 2009) seems to suggest that

sophisticated behavior by MPs is more likely than critics (e.g., Krehbiel and

Rivers, 1990; Groseclose and Milyo, 2009, 2010) have implied.

If sophisticated behavior is, however, part of reality, understanding its ex-

tent and its role is crucially important. Thus, in the present paper we present

evidence for strategic behavior by MPs relying on a novel strategy. While pre-

vious work on strategic behavior has tried to infer sincere preferences from the

roll call vote record (e.g., Bjurulf and Niemi, 1978; Denzau, Riker and Shep-

sle, 1985; Calvert and Fenno, 1994; Leemann, 2009) or proxies for preferences

(e.g., Enelow and Koehler, 1980; Enelow, 1981), we propose to employ direct

preference measures obtained through an MP survey. This allows us to demon-

strate with much more accuracy whether MPs behaved strategically and to what

extent. In addition, since our empirical data comes from a parliament where

party discipline is stronger than in the US Congress, we also consider how both

constituency and party interests affect the MPs’ decisions in strategic situations

(see for instance Denzau, Riker and Shepsle, 1985).

In the remainder of the paper we proceed as follows. In the next section

we briefly review work on strategic voting in parliaments. Since the possibility

1It might be that this sincere behavior is actually the result of sophisticated calculations, asin Austen-Smith’s (1987) model. See also Roberts (2007) for evidence on how strategic votingaffects ideal-point estimations.

2

for sophisticated voting strongly depends on institutional details, as Krehbiel and

Rivers (1990) convincingly argue, we present in section three a detailed account of

how lawmaking proceeds in the parliament on which we focus, namely the Swiss

lower house. In section four we present an analysis of responses to questions

in our MP survey, which allows us to assess whether MPs engage in strategic

thinking. We employ this information in our analysis of a decision for which we

have detailed preference information in section five. Section six concludes.

2 Strategic voting in parliaments

At least since Marquis de Condorcet’s (1785) work on majority rule and Arrow’s

(1951) study on social choice more generally the possibly strategic behavior of

individuals in collective choice situations has interested scholars. Riker (1958)

offered an early example and the seminal work by Farquharson (1969) presented

innovative ways how to analyze strategic voting, while the Gibbard-Satterthwaite

theorem (Gibbard, 1973; Satterthwaite, 1975) demonstrated the ubiquitous na-

ture of manipulation in voting. The theoretical work on strategic voting spurred a

series of empirical analyses, for instance by Bjurulf and Niemi (1978) and Enelow

and Koehler (1980). This early work on strategic voting in parliaments essentially

focused on situations where more than two options were under consideration and

attempted, on the basis of the roll call votes, to infer sophisticated behavior by

MPs. Relying essentially on Farquharson’s (1969) approach the idea was to iden-

tify the agenda tree and under the assumption of complete information on the

preferences of the MPs to assess whether strategic voting, i.e., voting against

the MPs’ sincere preferences, occurred (or was likely to have occurred) at one or

several nodes of the agenda tree.2

As Enelow (1981) in his extension of Farquharson’s (1969) theory rightly ar-

gued, complete information on the MPs’ preferences is hardly the typical situation

in a parliament. Thus, by allowing for private probabilistic assessments of the

outcomes of particular votes, he was able to demonstrate that strategic voting

depends essentially on two elements. On the one hand the probabilistic assess-

ments over the preferences of the MPs, and on the other hand the MP’s utility

2See McKelvey and Niemi (1978) for a discussion of Farquharson’s (1969) approach and thepresentation of a simpler procedure to determine whether strategic behavior occurs based onthe notion of “strategic equivalents.”

3

for the second best alternative.3

While a series of authors (Riker, 1958; Bjurulf and Niemi, 1978; Enelow and

Koehler, 1980; Enelow, 1981) were able to demonstrate that some MPs behaved

strategically in specific votes, under the assumption that the agenda tree was

known to the MPs, some MPs, even though they had an incentive for doing so,

failed to behave strategically. Denzau, Riker and Shepsle (1985) propose as an

explanation for this “deviant” behavior that for some MPs voting strategically

might be too costly in terms of their standing in their constituency (thus the

reference in Denzau, Riker and Shepsle’s (1985) title to Farquharson (1969) and

Fenno (1978)). Under the assumption of a known agenda tree, the authors show

both theoretically and empirically that strategic voting occurred, but that some

MPs refrained from such behavior because of, they argue, constituency prefer-

ences.

A much more sanguine view on sophisticated voting in Congress appears in

Krehbiel and Rivers (1990). These authors argue that to demonstrate the occur-

rence of sophisticated voting, the agenda tree must be fixed, and information on

preferences of the MPs widely known. According to them, these conditions were

not fulfilled in the cases discussed in Riker (1958) and Denzau, Riker and Shepsle

(1985). In addition Krehbiel and Rivers (1990) emphasize, based on Austen-

Smith’s (1987) work, that if the agenda is endogenous and the MPs’ preferences

are known to all, MPs will vote in each vote according to their sincere preferences,

but this is, as Austen-Smith (1987) labels it, “sophisticated sincerity.”4 Hence,

sophisticated voting is observationally equivalent to sincere behavior.

Calvert and Fenno (1994), relying on new examples, analyze a situation where

it is assumed that incomplete information is present both with respect to the

agenda and the MPs’ preferences. Finding evidence for sophisticated voting

by Congressmen, they challenge Krehbiel and Rivers’s (1990) assessment about

strategic behavior.

Most of the more recent work, however, takes up Krehbiel and Rivers’s (1990,

526, emphasis in the original) suggestion that “[t]he major outstanding issue is

not whether legislators are sophisticated but rather at what stages in the legisla-

3This second best alternative plays a crucial role in votes over three possible outcomes.4In a later study Groseclose and Krehbiel (1993) prove that “sophisticated sincerity” is

only possible when one particular agenda structure is used. More recently Groseclose andMilyo (2009, 2010) present a model where MPs, after voting, can switch their votes. In such asituation, sophisticated behavior is not part of equilibrium.

4

tive process their sophistication can be unambiguously observed and construc-

tively studied.” Londregan (2000) considers how committees structure voting in

the Chilean parliament and thus affect the agenda on the floor. Similarly Clinton

and Meirowitz’s (2003) theoretical foray suggests more complex analyses to assess

whether sophisticated voting is occurring (see Clinton and Meirowitz, 2004; Clin-

ton, 2006; Clinton, 2007).

Much of the literature on strategic behavior in parliaments deals, however,

with votes in the US Congress. Hence, not surprisingly much of the debate has

focused on whether sophisticated behavior is to be expected in a very specific

institutional context. This context, characterized by rather flexible agendas and,

comparatively speaking, weak political parties, is, however, rather atypical in the

broader world of representative democracies. Agenda control is quite different in

most other parliaments, as recent work clearly demonstrates (e.g., Doring, 1995;

Cox, Masuyama and McCubbins, 2000; Amorim Neto, Cox and McCubbins, 2003;

Chandler, Cox and McCubbins, 2006; Cox, 2006; Prata, 2006; Cox, McCubbins

and Skjaeveland, 2007; Cox, Heller and McCubbins, 2008; Brauninger and Debus,

2009).

Hence, even if the challenge of largely endogenous agendas raised by Krehbiel

and Rivers (1990) is certainly valid for the US Congress, few parliaments allow

for so much leeway in the formation of the agenda as is posssible on Capitol Hill.5

For instance, the empirical case we will discuss in more detail below, relies on a

variant of the so-called standard agenda (Miller, 1995, 13). In this variant, the

order in which amendments are pitted against each other is determined by their

content. Hence, the voting process only starts once all the amendments have been

proposed. The implication is that the agenda tree is known by the MPs once the

voting over competing amendments starts.6 Hence, a better understanding of

strategic behavior may be obtained when looking beyond the US Congress.

A second reason for looking beyond Congress is that while early studies (e.g.,

Bjurulf and Niemi, 1978; Enelow and Koehler, 1980; Enelow, 1981) looked at

strategic voting in isolation, it was only the path-breaking study by Denzau,

Riker and Shepsle (1985) that simultaneously considered sophisticated behavior

5See Miller (1995), Rasch (2000), and Schwartz (2008) (though see Miller, 2009) for a dis-cussion of the various types of agendas employed in parliaments.

6The official record of the two Swiss chambers is replete with statements by the chamberpresidents explaining and motivating the order in which amendments will be voted upon. Seehttp://www.parlament.ch/ab/frameset/f/index.htm.

5

by MPs and constituency effects. While these authors also mention in passing

the collective action problem of parties, it is not surprising that this factor is

not included in their analyses (even though based on the work by Cox and

McCubbins (2005, 2007), one would expect that deviations from strategic voting

might also be due to party pressure). Even if Cox and McCubbins (2005, 2007)

should be wrong, and parties hardly structure Congress, it would be foolish to

banish the effect of party pressure from studies of voting behavior in parliaments,

simply because very few parliaments have members which are as independent as

US Congressmen.7

Consequently, a better understanding of sophisticated voting by MPs might

be gained outside the US Congress, where in many parliaments the agenda is

much more structured and political parties affect more strongly the voting be-

havior of MPs (e.g., Carey, 2008).8 Hence, we propose in what follows a study of

sophisticated behavior in a parliament, where the agenda is much more tightly

set (which is more typical of most parliaments), and political parties affect con-

siderably the voting behavior of MPs.9

An additional innovation of the present study is that instead of relying on roll

call information or rather remote proxies to generate sincere preference orders over

alternatives, we rely on directly elicited preference orders from MPs. Proceeding

like this allows for a much more precise test of whether sophisticated voting occurs

or not. In addition, and more playfully, we have also tried to assess the strategic

capacities of MPs with the help of hypothetical questions (see section 4). Before

proceeding to these data matters we present, however, the institutional details of

the Swiss parliament.

7See for recent examples the work by Sieberer (2006), Calvo (2007) and Carey (2008).8See also Chandler, Cox and McCubbins (2006, 89), who argue that party discipline in most

parliamentary democracies is due both to agenda control and “whipping” of the MPs on thefloor.

9While several studies assess party discipline in Switzerland (e.g., Hertig, 1978; Luthi, Meyerand Hirter, 1991; Lanfranchi and Luthi, 1999; Bailer, Butikofer, Hug and Schulz, 2007; Hug,2010), Bailer, Butikofer, Hug and Schulz (2007) is the first study to disentangle party pressurefrom ideological similarity among party members.

6

3 Voting in the two chambers of the Swiss par-

liament

The voting procedure of the Swiss Parliament is regulated by the federal act on

the Federal Assembly. Also known as the “law on the parliament” (ParlG), it

was adopted on the basis of article 164, paragraph 1 of the Swiss Constitution

and came into effect on December 1, 2003. Figure 1 shows the sequence of

the lawmaking process in Switzerland. Our study focuses on the parliamentary

phase (stage two), in which the following voting procedure is employed.10 If

two mutually exclusive amendments are proposed regarding one article or one

paragraph of an article, they are put to a vote against one another. Should there

be more than two such amendments on a given article or paragraph, they are

voted on in the following order: First, the two most similar amendments are

put to a vote in order to choose among them. Then, the winning amendment

is set against the next most similar proposal. This procedure goes on until a

single amendment remains, which is then pitted against the status quo in the

last vote (von Wyss, 2001). Thus, the two chambers of the Swiss parliament

use a variant of what Miller (1995, 13) labels as standard agenda.11 Proposals,

most of which are initiated by government, are debated first in committee before

the deliberations start in one of the houses.12 The first deliberating chamber

begins by voting on whether the debate on the bill is accepted or not. If this

vote is positive, the plenary deliberation starts. Articles and paragraphs that

are undisputed are not submitted to a vote. All amendments are subject to

votes, following the procedure described above. The first house also concludes

its deliberation with a final vote. Final votes in the lower house are roll call

votes and the individual voting behavior of each MP is published in the Official

Bulletin.13

10The same procedure is also used in parliamentary committees.11In the basic form of the standard agenda (as used for instance in Congress) amendments

are voted upon in the order in which they are proposed (see also Rasch, 2000; Schwartz, 2008).This, as Krehbiel and Rivers (1990) forcefully argue, obviously allows for the introduction ofnew amendments when the voting has already started. Their critique of work on sophisticatedvoting in Congress is mostly based on this feature of the standard agenda.

12There is no fixed rule stating whether the upper or lower house starts the deliberation.This is decided for each bill by a joint committee (the “Ratsburo”) including among others thepresidents and vice-presidents of the two chambers.

13See Hug (2010) for a discussion of the circumstances under which roll call votes occur inthe Swiss parliament, and what information is publicly available.

7

Upon completion in the first chamber, the deliberation starts in the committee

of the second chamber. This committee discusses the current version of the bill,

that is, the version submitted to the final vote in the first house. The committee

can review all the amendments accepted during the previous deliberation and it

can decide to eliminate them. After the final vote in the committee, the plenary

discussion starts in the second chamber, again following the same procedure as

the one presented above. This deliberation concludes with a final vote, too. If

both chambers agree on a new bill, a final passage vote is taken in both of them.

Figure 1: Law making process in Switzerland

8

Should the versions adopted by the two houses differ, the bill goes back to

the first house, initiating a second phase of deliberation. The aim of this second

deliberation is to reach an agreement. If there is still no agreement between

the two houses after three such phases of deliberation, a conciliation committee

(“Einigungskonferenz”) is formed. This assembly, composed of members of both

houses, tries to solve the differences of opinion and proposes a new version of the

bill. This proposition is then submitted to a vote in each of the two houses. The

bill fails as a whole, if the proposition of the conciliation committee is rejected

by any of the houses in this vote.

4 Strategic thinking by MPs

While Denzau, Riker and Shepsle (1985) in their study tried to explain why some

congressmen with clear incentives for sophisticated behavior voted sincerely with

constituency considerations, it might also be that MPs differ in their ability to

perceive the opportunity for such behavior. To assess this possibility, we asked

in our MP survey the following two questions:14

You are supporting a project that entails expenditures of 100 millions

Swiss Francs (maximal). Lower expenditures would in your opinion

endanger the realization of the project. Two amendments are in-

troduced that would reduce the expenditures to 50 millions (median

proposal), respectively 30 millions (minimal proposal). In a first vote

the median proposal (50 millions) and the minimal proposal (30 mil-

lions) are brought to a vote. Do you vote for the median proposal (50

millions) or the minimal proposal (30 millions)?

Before the votes on these three proposals you are told that 99 mem-

bers of the lower house (resp. 22 members of the upper house) (the

president abstains) prefer the median proposal (50 millions) to the

maximal proposal (100 millions) and the latter (100 millions) to the

minimal proposal (30 millions). In a first vote the median proposal

14The MP survey was carried out between late 2006 and early 2007 with a response rateof approximately 60 % in the two chambers (out of 200, respectively 46 MPs). Most of theresponses were obtained online, while the remainder were obtained either by mail or by personalinterviews (see for more details Bailer, Butikofer, Hug and Schulz, 2007; Bailer and Butikofer,2009).

9

(50 millions) and the minimal proposal (30 millions) are brought to

a vote. Do you vote for the median proposal (50 millions) or the

minimal proposal (30 millions)?

In the second question the MP is informed that s/he is the pivotal voter in

both the first vote and the second vote. Thus by voting strategically for the

minimal proposal in the first vote s/he can induce a vote between the minimal

proposal and the maximal proposal in which s/he would again be the pivotal

voter and thus be able to assure that his/her preferred proposal wins. Table

1 cross-tabulates the responses to these two questions for the MPs of the two

houses separately.

Table 1: Strategic thinking of MPs (percentages (n in parentheses))vote intention after information

vote intention minimal medianbefore information (30 mio) (50 mio) abstention no response total

upper houseminimal (30 mio) 33.33 66.67 0.00 0.00 11.11

(1) (2) (0) (0) (3)median (50 mio) 12.50 81.25 0.00 6.25 59.26

(2) (13) (0) (1) (16)abstention 0.00 0.00 33.33 0.00 11.11

(0) (0) (1) (0) (3)no response 0.00 40.00 0.00 60.00 18.52

(0) (2) (0) (3) (5)total (27)

lower houseminimal (30 mio) 89.47 10.53 0.00 0.00 15.32

(17) (2) (0) (0) (19)median (50 mio) 11.90 79.76 2.38 5.95 67.74

(10) (67) (2) (5) (84)abstention 15.38 46.15 30.77 7.69 10.48

(2) (6) (4) (1) (13)no response 0.00 0.00 0.00 100.00 6.45

(0) (0) (0) (8) (8)total (124)

To the first question a large majority of MPs (59.26 % in the upper house

and 67.74 % in the lower house) responded to be voting for the proposal with

the higher expenditure level. Of these respondents only about a tenth (12.5 %

10

in the upper house and 11.9 % in the lower house) change their voting intention

in a strategic way after having obtained the information about the likely votes of

their colleagues and switch their vote. Interesting to note is that among the MPs

of the lower house almost 15 % affirm wanting to vote for the minimal proposal in

both votes, while this is only the case for approximately 3 % in the upper house.

Before turning to the real strategic behavior of MPs we present responses to

another question in our MP survey, namely whether MPs obtain information on

the likely vote distribution before votes over several amendments. 76 % of our

respondents mention that before such votes they obtain information on the likely

vote share for the various options. The frequency of such information according to

these 86 MPs (out of 113 having responded to this question) is depicted in figure

2. The figure clearly shows that MPs are often informed, for instance by their

party’s leadership, on the expected vote distribution in more complicated voting

processes. More than half of the MPs who stated they got information on the vote

shares to be expected estimated that frequency to be at least every second vote,

if not more. Hence, while complete information on all MPs sincere preferences

is unlikely to hold in most parliamentary settings, it is still very plausible that

MPs have a very good sense about the vote distribution to be expected, thanks

not the least to party leaders.

11

Figure 2: Frequency of information made available to MPs on likely vote shares

number of times (per ten) expected votes communicated to MP

Per

cent

of T

otal

0

5

10

15

20

1 2 3 4 5 6 7 8 9 10

12

5 Strategic voting in an agenda tree

Following the tradition of previous studies of strategic voting in parliaments (e.g.,

Riker, 1958; Bjurulf and Niemi, 1978; Enelow and Koehler, 1980; Enelow, 1981;

Denzau, Riker and Shepsle, 1985; Krehbiel and Rivers, 1990; Calvert and Fenno,

1994) we focus on individual bills.15 From our MP survey we have preference

orders over a series of alternatives for two bills debated in the Swiss parliament

between 2004 and 2007.

5.1 High speed railway network

The first bill deals with the connection of Swiss railroads to the trans-European

high-speed trains railway network in neighboring countries. For this bill we have

the preference order over four proposals:

• maximalist proposal: proposal by the committee majority

• intermediary proposal 1: proposal on the floor by MP Walker

• intermediary proposal 2: proposal by the committee minority (proposed by

MP Giezendanner)

• minimalist proposal: proposal on the floor by MP Fohn

Figure 5.1 shows how the MPs voted on these proposals according to the logic

discussed in the section 3, i.e., starting by voting on proposals deviating the most

from the committee proposal and being most similar to each other.16

15See Ladha (1994), Wilkerson (1999), Jenkins and Munger (2003) Finocchiaro and Jenkins(2008), and Leemann (2009) for different attempts to identify possible votes with strategicvoting.

16These votes occurred in December 2004 with the final passage vote in March 2005. Giventhat our MP survey was carried out after these votes, there might obviously be some contam-ination. The results reported below suggest that if such contamination occurred, it is ratherminor. In addition for our second bill most of the votes occurred after our MP survey, so thata comparison of the two cases allows us to assess this contamination in more detail.

13

Fig

ure

3:A

genda

tree

for

the

vote

onth

eco

nnec

tion

toth

ehig

h-s

pee

dtr

ains

railw

aynet

wor

k(L

egen

d:D

ashed

lines

indic

ate

the

bra

nch

es(o

pti

ons

inb

old)

chos

enby

the

MP

s,w

hile

the

num

ber

sat

the

nodes

indic

ate

the

tally

ofth

evo

tes.

)

min

imalist

proposal

interm

edia

ry

proposal2

1181:5

interm

edia

ry

proposal1

interm

edia

ry

proposal2

267:1

18

interm

edia

ry

proposal1

min

imalist

proposal

2

max.proposal

i.p.2

3m

ax.proposal

i.p.1

391:9

2m

ax.proposal

min

.p.

3m

ax.proposal

i.p.1

3

1

14

Table 2 reports the sincere preferences derived from the preference orderings

for all potential votes that might have occurred under the agenda tree depicted

in figure 5.1. In parentheses appear the actual results in parliament for the votes

that have occurred.

Table 2: Expected votes according to preferences (connection to the high-speedtrains railway network): row against column (in parentheses actual result inparliament)

optionsintermediary 1 intermediary 2 minimalist

maximalist option 53:58 69:37 78:32(92:91)

intermediary option 1 76:33 85:26(118:67)

intermediary 2 78:26(181:5)

Table 2 shows a few interesting elements. First, the intermediary option 1, i.e.,

the proposal of the government beats according to our preference information all

other options in pairwise comparisons. In the actual vote pitting this Condorcet

winner against the maximalist proposal, however, the latter won by a margin of

92 to 91 in the lower house of parliament.17

Second, the minimalist option was clearly a Condorcet loser, since all other

options debated were preferred by clear majorities to it. Hence, it cannot surprise

that in the first vote where this option was voted upon, it fared so badly. Of the

approximately 200 MPs present only five voted for it when it was pitted against

the intermediary option 2. As table 2 shows, more than five-times as many MPs

had expressed a preference for this option. Hence, at least 20 MPs voted against

their preferences in the first stage of voting.

5.2 CO2 incentive tax

The second bill for which we have detailed preference information concerned the

introduction of a CO2 incentive tax. We first depict in figure 5.2 part of the

agenda when the MPs considered article one of the CO2 incentive tax law. We

17Note that also the vote margin in our preference information from our MP survey is rathernarrow.

15

omitted a first vote over two minority proposals, namely the one of minortiy

1 wishing a much more radical increase in taxes which was pitted against the

proposal of minority 2, namely the government proposal. This latter proposal

won by a sizeable margin (107 yea against 68 nay).

The second element we obmit from the agenda in figure 5.2 is the very last

vote which concerned a proposal by minority 5. This proposal was only proposed

in case the proposal of minority 4 should fail (which it did). For both of these

elements we fail to have information on the preferences of the MPs. Given that

they concern the very first stage and a conditional amendment, our analyses

should not be affected too strongly.18

With these omissions the agenda depicted in figure 5.2 shows how a series

of minority proposals were pitted against the proposal from the majority of the

committee. All these proposals failed so that the committee majority got its most

prefered outcome.

The options for which we have preference information are the following

• majority proposal: majority of committee proposal

• minority 2: government proposal

• proposal Wafler: climate cent

• minority 3: link with fuel prices

• minority 4: status quo with voluntary measures

In table 3 we report for all options considered during the voting on Arti-

cle 1 the preference relationships as reported in our MP survey, as well as the

corresponding voting results.

18Especially the last omitted element should not be too problematic, since MPs should votesincerely in this last vote. This is also what we will assume in what follows.

16

Fig

ure

4:A

genda

tree

for

the

vote

onth

eC

O2

ince

nti

veta

x(L

egen

d:D

ashed

lines

indic

ate

the

bra

nch

es(o

pti

ons

inb

old)

chos

enby

the

MP

s,w

hile

the

num

ber

sat

the

nodes

indic

ate

the

tally

ofth

evo

tes.

)

min

ority

3proposal

majo

rity

proposal

2100:8

1

min

ority

2proposal

majo

rity

proposal

3147:3

0

majo

rity

proposal1

min

ority

3proposal

3

Wafler

majo

rity

proposal4

155:5

Wafler

m.2

p.

4W

afler

m.3

p.

4W

afler

m.p.

4

m.4

p.

m.p.

5110:7

0m

.4p.

W.

5m

.4p.

m.2

p.

5m

.4p.

W.

5m

.4p.

m.3

p.

5m

.4p.

W.

5m

.4p.

m.p.

5m

.4p.

W.

5

1

17

Tab

le3:

Exp

ecte

dvo

tes

acco

rdin

gto

pre

fere

nce

s(C

O2

ince

nti

veta

x):

row

agai

nst

colu

mn

(in

par

enth

eses

actu

alre

sult

inpar

liam

ent)

opti

ons

min

orit

y2

Wafl

erm

inor

ity

3m

inor

ity

4m

inor

ity

5m

ajo

rity

52:5

465

:44

69:3

673

:39

?(3

147:

30)

(415

5:5)

(210

0:81

)(5

110:

70)

(610

7:75

)m

inor

ity

1?

??

??

(168

:107

)m

inor

ity

2(g

over

nm

ent

pro

pos

al)

63:4

769

:37

75:3

8?

pro

pos

alW

afler

(clim

ate-

cent)

52:5

272

:36

?

min

orit

y3

(lin

kw

ith

fuel

pri

ces)

72:3

4?

min

orit

y4

(sta

tus

quo)

?

min

orit

y5

(ma

jori

typro

pos

alplu

slink

wit

hfu

elpri

ces)

?

18

Table 3 reveals that the government proposal submitted by minority 2 was,

according to our preference information, a Condorcet winner, though again a

narrow one. Only a slight majority preferred the government proposal to the

proposal of the committee majority. And in the end the lower chamber adopted

the latter proposal. Also, the minority 4 proposal, which corresponds to the

status quo, was a clear Condorcet loser.

5.3 Sincere preferences and preferences over sophisticatedequivalents

To more clearly illustrate the link between sincere preferences and actual votes we

present the following set of tables, first dealing with the bill on the connection to

the high-speed trains railway network and then the one on the CO2 incentive tax.

For each of the two bills we discuss the votes in reverse order, since the strategic

equivalents are obviously determined by the expected votes in later stages of the

agenda tree.

Table 4 shows surprisingly that in this last vote on the former bill not all MPs

voted according to their preferences. One explanation for this might obviously

be party and constituency pressure as suggested by Denzau, Riker and Shepsle’s

(1985) work. Another is linked to the fact that after the votes in the lower house,

the upper house still had to vote on this bill.

Table 4: Preferences and votes (connection to the high-speed trains railway net-work): final vote

Preference for optionsVote for options intermediary 2 maximalistintermediary 2 32 11maximalist 20 33

In table 5 we show in the first half the same relationship between the sincere

preferences and the actual vote for the second vote. According to the MPs’

responses about a quarter of the MPs, for whom we have the preference orderings,

voted against their sincere preferences. Given that the maximalist proposal was

clearly preferred to the intermediary proposal 2, voting for the latter proposal was

implicitly a vote for the maximalist proposal, that is its strategic equivalent. We

19

depict the relationship between these preferences over sophisticated equivalents19

and the actual vote in the second half of table 5.20

Table 5: Preferences and votes (connection to the high-speed trains railway net-work): second vote (in parentheses MPs with preferences for both comparisons)

Preference for optionsVote for options intermediary 1 intermediary 2 intermediary 1 maximalistintermediary 1 57 (57) 10 (10) 39 (36) 31 (31)intermediary 2 11 (11) 18 (17) 6 (5) 23 (23)

Interesting to note is that the preferences regarding the maximalist proposal

against the intermediary proposal 1 better explain the vote for the intermediary

proposal 2 than the sincere preferences. On the other hand, many MPs preferring

the maximalist proposal over the intermediary proposal 1 nevertheless voted for

the latter proposal, despite the danger that this option would beat their preferred

proposal in the last round of voting. Given our information over the preferences

and the result of the final passage vote implies that a safer strategy would have

been to vote for the intermediary proposal 2, as this proposal was much more

likely to be defeated by the commission proposal (i.e., the maximalist proposal).

This surprising result might largely be due to the fact that the outcome in the

final vote was very narrow, which also transpired in our MP survey.

This general analysis begs the question how MPs voted who preferred the

maximalist proposal to the intermediary proposal 1, which they preferred to the

intermediary proposal 2. Exactly these MPs should most heavily be tempted

to vote strategically for their least preferred option in the second to last vote.

Overall there are 37 MPs in this situation. Of these, however, only four voted

for their least preferred option in the second to last round.21

19To determine these strategic equivalents we rely exclusively on the information from ourMP survey. In the case of the final vote this leads to a situation where the strategic equivalent(intermediary proposal 1) does not correspond to the actually chosen alternative (maximalistproposal).

20In parentheses we report the numbers for those MPs only that appear also in the first partof the table.

21Interesting to note is that one out of these four had correctly indicated his strategic behaviorin the hypothetical questions discussed above in section 4. Among the 26 remaining MPs (forseven of the 37 MPs we have not the complete responses to the hypothetical questions) only 3(i.e. only 13 percent, against 25 percent) could be classified as sophisticated voters accordingto their responses to the hypothetical questions.

20

Table 6: Preferences and votes (connection to the high-speed trains railway net-work): first vote

Preference for optionsVote for options intermediary 2 minimalistintermediary 2 57 27minimalist 0 0

A similar phenomenon appears in table 6. As discussed above, the minimalist

proposal was, according to the preferences we elicited from MPs, a Condorcet

loser. Hence, voting for this option is clearly not optimal. No surprise then that

only 5 MPs voted on the floor for this option. Among the MPs that responded to

our survey we find, however, 27 MPs that preferred the minimalist option to the

intermediary option 2. None of these 27 MPs voted for their preferred option.

Strictly speaking in the first vote, the combination of our preference measure

and the actual results would suggest that either result in this first vote would lead

in the end to the adoption of the intermediary proposal 1. Hence, MPs should

have been indifferent between the two options if they considered the strategic

equivalents.22

The analysis along similar lines of the bill on the CO2 incentive tax is rendered

less interesting due to the fact that in all votes depicted in the agenda tree (see

figure 5.2) except the first vote the sophisticated equivalents coincided with the

options being voted upon.23 Nevertheless we depict in the following set of tables

again the relationship between the sincere preferences of MPs and their actual

voting behavior.

In the final vote on the one article of the CO2 incentive tax bill (table 7) we

find a rather close correspondence between sincere preferences and actual voting

decisions.24 Even stronger (and almost perfect) is this relationship in the second

to last vote (table 8). Here only a handful of MPs voted against their sincere

22If we consider, however, that given the closeness of the final vote the strategic equivalentfor voting for the intermediary proposal 2 is a vote for the maximalist proposal, we find only54 (46) MPs voting in such a sophisticated manner.

23In part this is due to the fact that we selected this bill before it was debated in all its detailin parliament. Hence, we expected sophisticated behavior but most likely the agenda structurereduced the likelihood of finding such behavior.

24In figure 5.2 this vote is not represented, since it occurred slightly after the votes depictedin the agenda tree and since it was largely a repeat of the last vote appearing in the figure.Their close relationship also easily transpires in tables 7 and 8.

21

Table 7: Preferences and votes (CO2 incentive tax): final passage vote (majorityvs. sq (minority 4))

Preference for optionsVote for options minority 4 majorityminority 4 33 4majority 3 67

preferences.

A more or less identical relationship we find for the second to last vote (which

is the last vote depicted in figure 5.2) (Table 8). Again a large majority of the

surveyed MPs voted according to their sincere preferences, and only a very small

minority voted against these preferences

Table 8: Preferences and votes (CO2 incentive tax): second to last vote (majorityvs. minority 4)

Preference for optionsVote for options minority 4 majorityminority 4 32 5majority 2 62

A rather puzzling result we find for the third to last vote, where the proposal

of MP Wafler attempted to eliminate the CO2 incentive tax and replace it by a

so-called climate cent. For this vote (Table 9) we find a large share of the MPs

voting against their preferred option (i.e. the proposal by MP Wafler). Most

likely at this stage the MPs already largely knew that the committee majority

proposal would win in the end.

Table 9: Preferences and votes (CO2 incentive tax): third to last vote (Waflervs. majority)

Preference for optionsVote for options Wafler majorityWafler 4 1 1majority 31 56

A similar picture transpires for the fourth to last vote depicted in table 10.

While almost all MPs who preferred the minority 2 proposal voted for this option,

22

among those MPs who preferred the majority proposal, a majority voted against

their sincere preferences. Nevertheless, this latter proposal passed with a large

margin (see figure 5.2). This is a rather puzzling finding, since according to our

preference measures the MPs should have considered this a vote where options

submitted corresponded to the strategic equivalents.

Table 10: Preferences and votes (CO2 incentive tax): fourth to last vote (minority2 vs. majority)

Preference for optionsVote for options minority 2 majorityminority 2 43 32majority 4 13

Finally, as discussed above, in the fifth to last vote, our preference informa-

tions would suggest a difference between sincere preferences and preferences over

strategic equivalents. We depict this in table 11. Disappointingly we find, how-

ever, a much stronger correspondence between sincere preferences and the actual

vote, than between preferences over strategic equivalents (second half of the table)

and the actual vote. As we will discuss below, however, whether voting according

to the preferences over strategic equivalents is attractive or not depends on the

more precise preference orderings.

Table 11: Preferences and votes (CO2 incentive tax): fifth to last vote (majorityvs. minority 3)

Preference for optionsVote for options minority 3 majority majority minority 2minority 3 26 (24) 6 (6) 25 (25) 5 (5)majority 4 (4) 57 (57) 23 (23) 42 (38)

5.4 Sincere and strategic voting

Following Denzau, Riker and Shepsle’s (1985) pioneering work, we now consider

whether deviations from strategic behavior by MPs are due to pressures from

the MPs’ principals. While Denzau, Riker and Shepsle (1985), based on Fenno’s

(1978) important work on Congressmen in their constituency, focused on the

23

latter’s effect,25 in parliamentary systems an additional principal, namely the

MP’s party, becomes also important.26

To assess whether deviations from strategic behavior are due to the preferences

of the MP’s principals we proceed as follows. Given that the votes on the various

proposals related to the high-speed rail connections and the CO2 incentive tax

had implications both in terms of expenditures and in terms of the protection of

the environment through public transportation and fuel savings, we presume that

the left-right position of the MP’s principals is a sufficiently good proxy for the

latter’s preferences.27 Hence we use on the one hand the left-right position of the

MPs’ parties as estimated by the expert-survey of Benoit and Laver (2006). The

left-right positions of the constituencies were gleaned from the mean response

in each canton to an ideological self-placement question (i.e., a left-right scale)

asked in the Swiss Household Panel (2005).28

Table 12 reports results of probit models estimated for the vote already re-

ported in table 4.29 The results first show that sincere preferences have a con-

siderable impact on the voting decision in this last vote. When controlling for

the preferences of the MP’s principals, however, this effect is reduced and we

find a strong party effect. This implies two things. First of all, the sincere pref-

25In their concluding section Denzau, Riker and Shepsle (1985) also address briefly the is-sue of parties in Congress, but this is done in much less rigorous way then their analysis onconstituency pressure.

26Obviously, strictly speaking, the Swiss political system is not a parliamentary but an“assembly-independent system” (Shugart and Carey, 1992, 26). As numerous studies haveshown, however, party discipline in the Swiss lower house is considerable (e.g., Hug, 2010).

27Given that weaker proxies lead to attenuated estimates of the causal effects, our results onconstituency and party pressure will be underestimated and thus biased against our hypothe-sized effect.

28Given that the sample of this survey selects randomly among all households, there is muchmore information from larger than smaller cantons. For this reason and following Park, Gelmanand Bafumi (2006) (see also Gelman and Hill, 2006; Lax and Phillips, 2009) we estimated amultilvel-model and used post-stratification to obtain improved estimates of the average left-right positions in the cantons.

29Strictly speaking the model estimated has multilevel characteristics, since MPs belong togroups by party and electoral district which have their proper explanatory variables. Estimatesof a multilevel model suggested, however, that the substantive conclusions would not be affectedby using this alternative empirical model. In addition, given the small number of cases andthe complex cross-classification structure, the assumptions implied by a classical multilevel arequite demanding and unlikely to apply. In addition, one might expect, given the strategicnature of the interactions taking place in the agenda tree, that models taking this into account(e.g., Signorino, 1999) might be more appropriate. We side, however, with Carrubba, Yuenand Zorn’s (2007) argument and demonstration that a correctly specified logit or probit modelallows to uncover the relevant effects more easily and as correctly.

24

erences of MPs in the last round of voting were largely similar in each of the

parties represented in the lower house. Second, the party pressure (and to a

much more marginal extent, voters’ pressures) affected much more strongly the

voting decision in this last vote than the sincere preferences of the MPs. Only

this can explain why the proportionate reduction in error (PRE) increases quite

dramatically from the first to the second model (from 28% to 43%).

Table 12: Preferences, party and voter pressure, and votes (connection to thehigh-speed trains railway network): final vote (probit)

Model 1 Model 2(Intercept) -0.33 1.19

(0.18) (1.91)sincere preference 0.98 ∗ 0.12

(0.27) (0.40)left-right position party -0.19 ∗

(0.04)left-right position voters 0.24

(0.39)n modal response 51 51n correctly predicted 64 74N 94 86AIC 120.25 79.31BIC 140.60 118.58log L -52.13 -23.65Standard errors in parentheses∗ indicates significance at p < 0.05

Table 13 reports similar analyses for the second vote depicted already in table

5. Again we find in the first two models strong effects for the sincere prefer-

ences, which remain even if we control for the preferences of two of the MPs’

principals. In the second model we find again a stronger effect for the MPs’

parties preferences than the preferences of the MPs’ voters. If we replace the

sincere preferences by the preferences over the sophisticated equivalents, we find

somewhat smaller coefficients for the preference variable30 and more importantly

smaller PREs. In models 5 and 6 we include both sincere preferences and pref-

30Please note that the preferences over the strategic equivalents are coded in the oppositeway, so that negative coefficients indicate a vote according to the preferences.

25

erences over sophisticated equivalents. The results suggest that both appear to

have affected the voting decision. The effect of the preferences over sophisticated

equivalents disappears, however, once we control for the preferences of the MPs’

principals. Hence, part of the sophisticated behavior of MPs might be due to

pressure by the MPs’ parties.

26

Tab

le13

:P

refe

rence

s,par

tyan

dvo

ter

pre

ssure

,an

dvo

tes

(con

nec

tion

toth

ehig

h-s

pee

dtr

ains

railw

aynet

wor

k):

seco

nd

vote

(pro

bit

)M

odel

1M

odel

2M

odel

3M

odel

4M

odel

5M

odel

6M

odel

7(I

nte

rcep

t)-1

.02

∗2.

37-0

.21

2.84

-0.6

7∗

4.46

1.90

(0.1

9)(2

.07)

(0.1

7)(2

.19)

(0.2

1)(2

.59)

(2.2

8)si

nce

repre

fere

nce

1.38

∗1.

77∗

1.40

∗1.

92∗

1.82

∗

(0.3

0)(0

.46)

(0.3

3)(0

.51)

(0.5

0)le

ft-r

ight

pos

itio

npar

ty0.

23∗

0.18

∗0.

20∗

0.23

∗

(0.0

5)(0

.05)

(0.0

6)(0

.05)

left

-rig

ht

pos

itio

nvo

ters

-1.3

5∗

-1.1

7∗

-1.7

0∗

-1.2

7∗

(0.4

9)(0

.46)

(0.5

7)(0

.51)

pre

fere

nce

over

sophis

tica

ted

equiv

alen

ts-0

.94

∗-0

.58

-0.9

3∗

-0.7

3(0

.31)

(0.4

6)(0

.34)

(0.5

3)st

rate

gic

ince

nti

ve0.

25(0

.54)

nm

odal

resp

onse

6666

6666

6666

66n

corr

ectl

ypre

dic

ted

7478

6673

8179

81N

9690

9385

9385

90A

IC96

.64

57.7

210

6.08

73.8

789

.06

57.7

059

.35

BIC

117.

1697

.72

126.

3411

2.95

119.

4510

6.55

109.

34lo

gL

-40.

32-1

2.86

-45.

04-2

0.93

-32.

53-8

.85

-9.6

7St

anda

rder

rors

inpa

rent

hese

s∗

indi

cate

ssi

gnifi

canc

eat

p<

0.05

27

These analyses already demonstrate that sophisticated behavior is well present

in this vote on the high-speed rail connections. Our indicator for preferences over

sophisticated equivalents does, however, not take into account that for some

MPs voting over sophisticated equivalents is not the optimal strategy in the

agenda tree.31 Hence, in model 7 we replace the preferences over sophisticated

equivalents by an indicator whether the MP in question had an incentive to

vote strategically for the intermediary proposal 2. As our results show, these

strategic incentives have a considerable effect of the voting behavior of MPs. The

substantive effect of this strategic incentive is illustrated in figure 5 which depicts

the simulated maximum effect of this variable, while holding all variables constant

at their means, except the preference variable which was set to its maximum

1.32 Interestingly this effect persists despite our controls for party and voter

preferences.

Figure 5: Density of maximum effect of strategic incentives (connection to thehigh-speed trains railway network): second vote

In table 14 we report for purposes of comparisons the results of a probit

model where the dependent variable is the final vote on the CO2 incentive tax

that pitted the status quo against the committee majority proposal.33 We find,

not surprisingly, strong effects for the sincere preferences and also a considerable

effect due to the ideological positioning of the MP’s party. The effect of the

voter’s preferences, however, is rather limited.

Given that in the whole sequence of votes for which we have preference infor-

mation sincere preferences correspond to preferences over strategic equivalents,

31We wish to thank Elisabeth Maggie Penn for suggesting this additional analysis.32We used the plot.zelig command of the Zelig package (Imai, King and Lau, 2008) to

produce this figure and figure 6.33The corresponding comparison between actual vote and sincere preferences appears in table

7.

28

Table 14: Preferences, party and voter pressure, and votes (CO2 incentive tax):final vote (probit)

Model 1 Model 2(Intercept) -1.54 ∗ 5.00

(0.34) (3.98)sincere preference 2.97 ∗ 2.25 ∗

(0.40) (0.71)left-right position party -0.46 ∗

(0.14)left-right position voters 0.07

(0.74)n modal response 70 68n correctly predicted 100 99N 103 94AIC 54.79 24.61BIC 75.87 65.30log L -19.40 3.70Standard errors in parentheses∗ indicates significance at p < 0.05

we refrain from presenting similar analyses as those reported in table 14 for these

votes. Only for the very first voted depicted in figure 5.2 are there differences

between the strategic equivalents and the options on the table. Hence, in table

5.4 we present the results of a model explaining the vote in this fifth to last vote.

The results echo largely those obtained for the first bill we analyzed. We find

a, though smaller, effect for the preferences over the sophisticated equivalents,

which, however, disappears completely once we control for the preferences of the

MP’s principals (model 4). More interesting, however, is again the effect for the

strategic incentive which we estimate in model 6. Again, beyond the sincere pref-

erences and the preferences of the MP’s principals we find a considerable effect

due to strategic calculations. We illustrate this effect again by depicting in figure

6 the maximum effect while holding constant all variables at their means except

the sincere preferences fixed at the value 1. As figure 6 nicely illustrates, the

probability of voting for the committee majority proposal increases substantially

if the MP’s preference ordering creates a strategic incentive to vote against his

or her sincere preferences.

29

Tab

le15

:P

refe

rence

s,par

tyan

dvo

ter

pre

ssure

,an

dvo

tes

(CO

2in

centi

veta

x):

fift

hto

last

vote

(pro

bit

)M

odel

1M

odel

2M

odel

3M

odel

4M

odel

5M

odel

6(I

nte

rcep

t)-1

.79

∗3.

60-0

.06

11.6

0∗

4.23

3.51

(0.5

1)(5

.02)

(0.2

9)(4

.91)

(5.0

8)(5

.22)

since

repre

fere

nce

3.95

∗2.

93∗

2.86

∗1.

70(0

.65)

(1.0

6)(1

.09)

(1.1

8)le

ft-r

ight

pos

itio

npar

ty-0

.57

∗-0

.78

∗-0

.63

∗-0

.63

∗

(0.1

6)(0

.18)

(0.1

6)(0

.17)

left

-rig

ht

pos

itio

nvo

ters

0.48

-0.1

60.

600.

72(0

.97)

(0.9

2)(1

.00)

(1.0

2)pre

fere

nce

over

sophis

tica

ted

equiv

alen

ts2.

08∗

-0.3

9-1

.02

(0.5

2)(0

.98)

(1.0

1)st

rate

gic

ince

nti

ve1.

53(1

.05)

nm

odal

resp

onse

6159

6563

5959

nco

rrec

tly

pre

dic

ted

8385

6589

8484

N95

8697

8885

85A

IC67

.26

31.8

110

2.38

36.9

631

.25

30.7

7B

IC87

.69

71.0

812

2.97

76.6

080

.10

79.6

3lo

gL

-25.

630.

09-4

3.19

-2.4

84.

374.

61St

anda

rder

rors

inpa

rent

hese

s∗

indi

cate

ssi

gnifi

canc

eat

p<

0.05

30

Figure 6: Density of maximum effect of strategic incentives (CO2 incentive tax):fifth to last vote (probit)

6 Conclusion

Strategic behavior by MPs occupies a rather ambiguous position in the literature

on parliamentary behavior. On the one hand it is disputed by some authors to

occur at all on the floor of parliaments (Krehbiel and Rivers, 1990; Groseclose

and Milyo, 2009, 2010), while others seem to find evidence that it occurs both

on the floor and in previous stages of a legislative process (most notably Calvert

and Fenno, 1994). On the other hand, despite this uncertainty whether strategic

behavior is important, a large part of the literature dealing with roll call votes, is

predicated on the assumption that MPs vote sincerely (for exceptions see Clinton

and Meirowitz, 2004; Clinton, 2007). Much of the discussion whether or not

sophisticated behavior by MPs can be observed focuses, however, on the US

Congress, and many of Krehbiel and Rivers’s (1990) and Groseclose and Milyo’s

(2009, 2010), critiques do not apply generally to parliaments around the world.

Starting from the fact that in most parliaments apart the US Congress the

agenda is much more fixed we proposed to study sophisticated behavior in a set-

ting where agendas hardly change in the course of voting on a bill. In addition, the

parliament we focused on, namely the Swiss parliament, is characterized both by

considerable party discipline and some constituency pressure (Hug, 2010). Hence,

the results we presented in this paper are likely to be much more representative of

parliamentary life elsewhere than studies on the US Congress. In addition, we re-

lied in our empirical assessment of strategic voting by MPs on a direct preference

measures gleaned from answers in a MP survey. This allows for a more precise

assessment of strategic behavior than what appears traditionally in studies on

this topic.

31

With these innovations we were able to demonstrate that strategic behavior

by MPs clearly occurs in the Swiss parliament. MPs voting in an agenda tree

seem quite well aware of the consequences of their votes in early stages and

behave accordingly. Our results also demonstrate that the effect of preferences

over sophisticated equivalents is somewhat mitigated by the preferences of the

MPs’ principals, namely the party to which they belong and their voters. Among

these two effects, the former is, however, considerably stronger.

These results should give some pause to scholars employing roll call data and

presume that this data is the expression of MPs’ sincere preferences. Even in

a situation where party pressure is considerable, strategic considerations affect

voting decisions by MPs. If this is the case, however, making sense out of roll

call data requires a much more detailed understanding of the legislative process

more generally. Hence, work following upon Clinton and Meirowitz (2004) and

Clinton (2007) seems thoroughly required.

32

Appendix

In table 16 we report the descriptive statistics of all variables employed in this

paper, except for the preference and voting variables, for which details appear in

tables 1-6 in the main text.

Table 16: Descriptive statisticsVariable Min Mean Max Std. Dev. n

MPs having participated in MP surveyleft-right position party 2.000 11.061 17.875 5.693 117left-right position voters 4.094 4.741 6.188 0.340 119strategist 0 0.091 1 0.289 110

33

References

Amorim Neto, Octavio, Gary Cox and Mathew D. McCubbins. 2003. “Agenda

Power in Brazil’s Camara Dos Deputados, 1989 to 1998.” World Politics

55:550–578.

Arrow, Kenneth J. 1951. Social Choice and Individual Values. New Haven: Yale

University Press.

Austen-Smith, David. 1987. “Sophisticated Sincerity: Voting over Endogenous

Agendas.” American Political Science Review 81(4):1323–1329.

Bailer, Stefanie and Sarah Butikofer. 2009. “Party Discipline in Switzerland.”

unpublished manuscript.

Bailer, Stefanie, Sarah Butikofer, Simon Hug and Tobias Schulz. 2007. “Prefer-

ences, Party Discipline and Constituency Pressure in Swiss Parliamentary

Decisions.” Paper prepared for presentation at the ECPR General Confer-

ence Pisa, September 6-8, 2007.

Benoit, Kenneth and Michael Laver. 2006. Party Policy in Modern Democracies.

London: Routledge.

Bjurulf, Bo H. and Richard G. Niemi. 1978. “Strategic Voting in Scandinavian

Parliaments.” Scandinavian Political Studies 1(1):5–22.

Brauninger, Thomas and Marc Debus. 2009. “Legislative agenda-setting in par-

liamentary democracies.” European Journal of Political Research 48(6):804

– 839.

Calvert, Randall L. and Richard F. Jr. Fenno. 1994. “Strategy and Sophisticated

Voting in the Senate.” Journal of Politics 56(2):349–376.

Calvo, Ernesto. 2007. “The Responsive Legislature: Public Opinion and Law

Making in a Highly Disciplined Legislature.” British Journal of Political

Science 36(2):263–280.

Carey, John M. 2008. Legislative Voting and Accountability. New York: Cam-

bridge University Press.

34

Carrubba, Clifford J., Amy Yuen and Christopher Zorn. 2007. “In Defense of

Comparative Statics: Specifying Empirical Tests of Models of Strategic In-

teraction.” Political Analysis 15(4):465–482.

Chandler, William, Gary W. Cox and Mathew D. McCubbins. 2006.

“Agenda Control in the German Bundestag, 1980-2002.” German Politics

15(March):89–111.

Clinton, Joshua D. 2006. “Testing Lawmaking Theories with (Endogenous) Roll

Calls, 90th - 106th U.S. House.” Department of Politics Princeton University.

Clinton, Joshua D. 2007. “Congress, Lawmaking, and the Fair Labor Standards

Act from 1971-2000.” Paper prepared to be presented at the Conference on

Measures of Legislators’ Policy Preferences and the Dimensionality of Policy

Spaces, Washington University in St. Louis, November 29- December 2.

Clinton, Joshua D. and Adam Meirowitz. 2003. “Integrating Voting Theory and

Roll Call Analysis: A Framework.” Political Analysis 11(4):381–396.

Clinton, Joshua D. and Adam Meirowitz. 2004. “Testing Explanations of Strate-

gic Voting in Legislatures: A Reexamination of the Compromise of 1790.”

American Journal of Political Science 48(4):675–689.

Cox, Gary W. 2006. The Organization of Democratic Legislatures. In The Ox-

ford Handbook of Political Economy, ed. Barry R. Weingast and Donald A.

Wittman. Oxford: Oxford University Press pp. 141–161.

Cox, Gary W. and Mathew D. McCubbins. 2005. Setting the Agenda: Respon-

sible Party Government in the U.S. House of Representatives. Cambridge:

Cambridge University Press.

Cox, Gary W. and Mathew D. McCubbins. 2007. Legislative Leviathan: Party

Government in the House. 2 ed. Cambridge University Press.

Cox, Gary W., Mathew D. McCubbins and Asbjorn Skjaeveland. 2007. “Agenda

Power and Lawmaking in the Danish Folketing.” Paper prepared for pre-

sentation in the Danish Politics Section, Department of Political Science,

University of Aarhus.

35

Cox, Gary W., Mikitaka Masuyama and Mathew D. McCubbins. 2000. “Agenda

Power in the Japanese House of Representatives.” Japanese Journal of Po-

litical Science 1(1):1–22.

Cox, Gary W., William B. Heller and Mathew D. McCubbins. 2008. “Agenda

Power in the Italian Chamber of Deputies: 1988 to 2000.” Legislative Studies

Quarterly 33(2):171–198.

de Condorcet, Marie Jean Antoine Nicolas Caritat Marquis. 1785. Essai sur

l’application de l’analyse de la probabilite des decisions rendues a la pluralite

des voix. Paris: Imprimerie Royale.

Denzau, Arthur, William H. Riker and Kenneth A. Shepsle. 1985. “Farquhar-

son and Fenno: Sophisticated Voting and Home Style.” American Political

Science Review 79(4):1117–1134.

Doring, Herbert. 1995. Time as a Scarce Resource: Government Control of the

Agenda. In Parliaments and Majority Rule in Western Europe, ed. Herbert

Doring. New York: St. Martin’s Press pp. 223–246.

Enelow, James M. 1981. “Saving Amendments, Killer Amendments, and an Ex-

pected Utility Theory of Sophisticated Voting.” Journal of Politics 43:1062–

1089.

Enelow, James M. and David H. Koehler. 1980. “The Amendment in Legislative

Strategy: Sophisticated Voting in the U.S. Congress.” Journal of Politics

5:396–413.

Farquharson, Robin. 1969. Theory of Voting. New Haven: Yale University Press.

Fenno, Richard F. 1978. Home Style: House Members in Their Districts. Boston:

Little, Brown.

Finocchiaro, Charles J. and Jeffrey A. Jenkins. 2008. “In Search of Killer Amend-

ments in the Modern U.S. House.” Legislative Studies Quarterly 33:263–294.

Gelman, Andrew and Jennifer Hill. 2006. Data Analysis Using Regression and

Multilevel/Hierarchical Models. New York: Cambridge University Press.

36

Gibbard, Alan. 1973. “Manipulation of Voting Schemes: A General Result.”

Econometrica 41:587–601.

Groseclose, Tim and Jeffrey Milyo. 2009. “Sincere Versus Sophisticated Voting

When Legislators Vote Sequentially.” Departments of Political Science and

Economics UCLA and Department of Economics University of Missouri.

Groseclose, Tim and Jeffrey Milyo. 2010. “Sincere Versus Sophisticated Voting

in Congress: Theory and Evidence.” Journal of Politics 72(1):60–73.

Groseclose, Tim and Keith Krehbiel. 1993. The Pervasiveness of Sophisticated

Sincerity. In Political Economy: Institutions, Information, Competition and

Representation, ed. William A. Barnett, Norman J. Schofield and Melvin J.

Hinich. New York:: Cambridge University Press pp. 247–278.

Hertig, Hans-Peter. 1978. “Party Cohesion in the Swiss Parliament.” Legislative

Studies Quarterly 3(1):63–81.

Hug, Simon. 2010. “Selection Effects in Roll Call Votes.” British Journal of

Political Science 40(1):225–235.

Imai, Kosuke, Gary King and Olivia Lau. 2008. “Toward A Common Framework

for Statistical Analysis and Development.” Journal of Computational and

Graphical Statistics 17(4):892–913.

Jenkins, Jeffery A. and Michael C. Munger. 2003. “Investigating the Incidence

of Killer Amendments in Congress.” Journal of Politics 65:498–517.

Krehbiel, Keith and Douglas Rivers. 1990. “Sophisticated Voting in Congress: A

Reconsideration.” Journal of Politics 52:548–578.

Ladha, Krishna K. 1994. “Coalitions in Congressional Voting.” Public Choice

78:43–64.

Lanfranchi, Prisca and Ruth Luthi. 1999. Cohesion of Party Groups and Inter-

party Conflcit in the Swiss Parliament: Roll Call Voting in the National

Council. In Party Discipline and Parliamentary Government, ed. Shaun

Bowler, David M. Farrell and Richard S. Katz. Columbus: Ohio State

University Press pp. 99–120.

37

Lax, Jeffrey R. and Justin H. Phillips. 2009. “How Should We Estimate Public

Opinion in The States?” American Journal of Political Science 53(1):107–

121.

Leemann, Lucas. 2009. “Catch me if you can. Recovering sophisticated voting

in the lower house of Switzerland.” Paper prepared for the Annual Swiss

Political Science Meeting, St.Gallen January 2009.

Londregan, John. 2000. Legislative Institutions and Ideology in Chile. New York:

Cambridge.

Luthi, Ruth, Luzius Meyer and Hans Hirter. 1991. Fraktionsdisziplin und die

Vertretung von Partiukularinteressen im Nationalrat. In Das Parlament -

“Oberste Gewalt des Bundes”?, ed. Parlamentsdienst. Bern: Haupt pp. 53–

71.

McKelvey, Richard D. and Richard G. Niemi. 1978. “A Multistage Game Repre-

sentation. of Sophisticated Voting for Binary Procedures.” Journal of Eco-

nomic Theory 18:1–22.

Miller, Kenneth P. 2009. Direct Democracy and the Courts. New York: Cam-

bridge University Press.

Miller, Nicholas R. 1995. Committees, Agendas, and Voting. Chur: Harwood.

Park, David K., Andrew Gelman and Joseph Bafumi. 2006. State Level Opinions

from National Surveys: Poststratification using Multilevel Logistic Regres-

sion. In Public Opinion in State Politics, ed. Jeffrey E. Cohen. Stanford,

CA: Stanford University Press.

Poole, Keith. 2005. Spatial Models of Parliamentary Voting. New York: Cam-

bridge University Press.

Prata, Adriana. 2006. Government Domination, Consensus or Chaos? A Study

of Party Discipline and Agenda Control in National Legislatures PhD thesis

University of California, San Diego.

Rasch, Bjorn Erik. 2000. “Parliamentary Floor Voting Procedures and Agenda

Setting in Europe.” Legislative Studies Quarterly 25(1):3–23.

38

Riker, William H. 1958. “The Paradox of Voting and Congressional Rules for

Voting on Amendments.” American Political Science Review 52:349–366.

Roberts, Jason M. 2007. “The Statistical Analysis Of Roll-Call Data: A Cau-

tionary Tale.” Legislative Studies Quarterly 32:341–360.

Satterthwaite, M.A. 1975. “Strategyproofness and Arrow’s Conditions: Existence

and Correspondence Theorems for Voting Procedures and Social Welfare

Functions.” Journal of Economic Theory 10:187–217.

Schwartz, Thomas. 2008. “Parliamentary Procedure: Principal Forms and Polit-

ical Effects.” Public Choice 136:353–377.

Shugart, Matthew Soberg and John M. Carey. 1992. Presidents and Assemblies.

Cambridge: Cambridge University Press.

Sieberer, Ulrich. 2006. “Party Unity in Parliamentary Democracies: A Compar-

ative Analysis.” Journal of Legislative Studies 12(2):150–178.

Signorino, Curtis S. 1999. “Strategic Interaction and the Statistical Analysis of

International Conflict.” American Political Science Review 93(2):279–297.

von Wyss, Moritz. 2001. “Maximen und Prinzipien des parlamentarischen Ver-

fahrens.” PhD Dissertation University of Zurich.

Wilkerson, John D. 1999. “”Killer” Amendments in Congress.” American Polit-

ical Science Review 93(3):535–552.

39