Party Cohesion in Westminster Systems

Inducements, Replacement and Discipline in the House of Commons,

1836–1910∗

Andrew Eggers† Arthur Spirling‡

April 5, 2013

Abstract

We consider the historical development of a characteristic crucial for the functioningand normative appeal of Westminster systems: cohesive legislative parties. To do this,we gather the universe of the twenty thousand parliamentary divisions that took placebetween 1836 and 1910 in the British House of Commons, and identify whips therein.Using this new data, we construct a voting record for every Member of Parliament serv-ing during this time, and compile covariate information for the same. In line with pre-vious work, we show that—with the exception of a chaotic period in the 1840s—mediandiscipline was always high and increased throughout the century. We use novel ‘relativedistribution’ methods to provide more nuanced findings than previously possible, andshow that almost all the rise in cohesion results from the reduction of a rebellious ‘lefttail’, rather than central tendency shifts. In explaining the aggregate trends, we findstrong support for ‘inducement’ theories that stress the importance of internal organi-zation changes, and scant evidence for ‘replacement’ explanations which imply that newintakes of members were more disciplined than those leaving the chamber.

∗First version: March 2013. This version: April 5, 2013. Kathy Wang and Tania Amarillas provided excellentresearch assistance. We are grateful to the History of Parliament Trust for data. IQSS contributed financially to thecosts of this project, and its assistance is gratefully acknowledged.†London School of Economics. email: a.c.eggers@lse.ac.uk‡Department of Government, Harvard University. email: aspirling@gov.harvard.edu

1

1 Introduction

Strong party discipline is a defining feature of legislative politics in Westminster systems (Lijphart,

1999; Spirling and McLean, 2007; Kam, 2009b). In modern times, parliamentary divisions (‘roll

calls’ in American parlance) are characterized as predictable affairs insofar as members of parlia-

ment (MPs) can be expected to vote in line with their parties’ explicit orders, and rebellion against

the ‘whip’ is a relatively rare event (see Cowley, 2002, for an overview). One consequence, along

with an electoral system that yields a ‘manufactured majority’ of seats (Bogdanor and Butler,

1983), is that governing parties almost never lose votes important to their legislative agenda. This

arrangement facilitates the particular vision of “responsible” government (Birch, 1964; Gamble,

1990; Campbell and Wilson, 1995; Richards and Smith, 2002) and “accountability” (in the sense

of Powell, 2000) that has allowed unusual longevity and diffusion of the Westminster system from

its home in Britain to other polities (Rhodes and Weller, 2005; Rhodes, Wanna and Weller, 2009).

Given the centrality of cohesive parties to both the functioning of the Westminster system and

to the normative justification of its use, it is unsurprising that a great deal of political science at-

tention has been lavished upon them. Starting as early as Lowell (1902), scholarship has focussed

on the 19th Century as a transitional time during which party cohesion—and thus the Westminster

system—took hold in the House of Commons (e.g. Bagehot, 1873/2011; Trevelyan, 1922; Aydelotte,

1963; Berrington, 1968; Cromwell, 1982; Cox, 1987; Jenkins, 1996; Rush, 2001; Schonhardt-Bailey,

2003). While researchers have come to a clear consensus that discipline did indeed increase, the

central question of why MPs became more cohesive has lacked a convincing answer. Though we

are not without theories (with Cox, 1987, being perhaps the most comprehensive), demonstrat-

ing conclusively which particular mechanism explains the patterns we see has remained an elusive

goal. To the extent that understanding parliamentary discipline in modern times is an important

goal in political science (e.g. Norton, 1978; Diermeier and Federsen, 1998; Cowley, 2005; Benedetto

and Hix, 2007) and to the extent that modern politics relies upon it (e.g. Economist, 2012), there

is obvious interest in improving on this state of affairs. This paper represents such an improvement.

2

At its core, the question is this: did parties become more cohesive primarily because MPs who

were previously less disciplined became more disciplined as the century progressed; or was the in-

crease in cohesion due to more disciplined ‘types’ of individuals entering the Commons while less

disciplined ‘types’ left it? The former story implies that change was mostly due to factors acting

on already existing members—like increasingly efficient whipping, a greater sense of solidarity, or

responsiveness to a party orientated electorate. We denote this an ‘inducements’ account. The

latter story suggests that it is newcomers to the chamber who brought with them cohesion, while

a rebellious ‘old guard’ were replaced by natural wastage. We denote this a ‘replacement’ account.

Determining which of these mechanisms matters—or in what proportion they matter—is as funda-

mental as it difficult. That previous work has been unable to decisively resolve the issue is in part

due to methodological shortcomings, but mostly due to issues of data availability. Here, we solve

both problems using novel statistical methods and a new collection of roll call records.

Our data consists of every one of the 20262 roll calls cast between 1836 and 1910. We have a

record of every member’s decision on every division. By combining this voting information with a

new data set that incorporates extensive individual-level covariates on the thousands of MPs serv-

ing during this time, we are able to explore discipline in more nuanced ways than previous studies

that used aggregated statistics (e.g. Lowell, 1902; Berrington, 1968; Cox, 1987) ever could. Simi-

larly, we are able to chart the changes to cohesion over a much broader period than the intensive,

but time-limited studies (e.g. Adyelotte, 1954; Cromwell, 1982; McLean, 2001; Schonhardt-Bailey,

2003) preceding ours. In short, we have the best of both worlds: micro-data as a long time series.

Our statistical approach is to make use of ‘relative distribution methods’ (Handcock and Morris,

1998, 1999) which allow us to compare entire ranges of discipline between parliaments without

having to resort to typically misleading aggregate summaries like means or standard deviations.

Our chosen technique is non-parametric, and robust to outliers (a common problem in histories of

legislative voting).

Starting from a high baseline, we find that discipline in parliament waned in the 1840s, and then

3

grew thereafter. This observation is commensurate with earlier accounts, though we are able to be

much more precise in our discussion of changes. Importantly, we show that cohesion was always

high in absolute terms for the average MP, and that almost all growth in aggregate party-line voting

is due to the disappearance of a rebellious ‘left tail’ from the distribution of discipline. We then

show that this ‘folding in’ of the rebel tail is not due to new blood entering the House of Commons

or from older less-malleable types retiring out: instead, we demonstrate that sitting MPs became

more cohesive as the Victorian age wore on, suggesting that it is party organization and other

‘endogenous’, internal reforms and incentives that mattered most (in the sense of e.g. Cox, 1987;

Kam, 2009b), rather more sociological change to MP stock (in the sense of e.g. Rush, 2001).

Our paper continues as follows: in Section 2 we review the literature pertinent to the question

at hand, and set up our empirical task. In Section 3 we introduce our new data, and in Section

4 we introduce methods for dealing with them. Section 5 explores our results, while Section 6

concludes.

2 Literature and Orientation

The period between the first Reform Act of 1832, and the fourth Reform Act of 1918, has been

cited by scholars as the key period of development for the both the Westminster system, and the

British polity as a whole (e.g. Cox and Ingram, 1992; Jenkins, 1996; McLean, 2001; Rush, 2001).

Central to this story is increased party cohesion in legislative voting, and scholars have expended

considerable efforts in measuring it (e.g. Lowell, 1902; Berrington, 1968). Though different authors

approach the specific measurement problems in varying ways, (Cox, 1987, 26) summarizes the con-

sensus finding of the literature when he notes that “For the Conservatives, there is a sharp decline

[in the 1830s and 1840s]. . . a plateau in the 1850s; and a sharp recovery [later]. . . ” Meanwhile, “[f]or

the Liberals, there is a much smaller decline. . . followed by an erratic increase”.

Exactly why cohesive parties became increasingly the norm is less clear cut. Understanding the

pattern obviously requires understanding how MPs both sorted themselves into parties, and why

4

they behaved the way they did once sorted. Focussing on the 1840s—an unusually chaotic time

in British legislative history—some scholars have emphasized material interests as the drivers of

MP-behavior (Adyelotte, 1954; Schonhardt-Bailey, 2003), while others have sought explanations

in ideology (McLean, 2001). Related literatures point to electoral reform as ushering in the “tri-

umph of partisan politics” in the later Victorian period, with polarization of the electorate having

knock-on consequences for the House of Commons itself (Jenkins 1996, see also Cox and Ingram

1992). With a view to the broader time period and more attention to internal reform, Cox (1987)’s

account of discipline—in which ministerial promotion is the valued reward for a career of obedi-

ence to a party in parliament—is perhaps the best known account that links external changes to

suffrage and individual motivations for division behavior. More recently, scholars have researched

changes in the personal socioeconomic circumstances of MPs during this period, though without

particular attention to ‘interests’ per se: Rush (2001), for example, notes the pronounced decline in

members with aristocratic links as the century wore on, and the commensurate rise of businessmen

and representatives of the (legal) professions. Connected to such reasoning is a broader literature

emphasizing the importance of campaign finance arrangements in a more modern democratic age

(Hanham, 1978; Kam, 2009a; Stokes, 2011), and the need to avoid falling into bad odor with a

central party administration.

Clearly, there were internal and external pressures for MPs to consider when taking part in di-

visions in the House of Commons. And these pressures waxed and waned over time. Still, whatever

subset of theories one subscribes to for explaining the rise in discipline, Krehbiel (1993)’s critique

most assuredly applies; that is, from roll call votes alone it is impossible to know whether MPs fol-

lowed the party line because they faced inducements to do so, or whether they would have done so

anyway due to their latent ideological positions (and thus party membership is simply a by-product

of less malleable characteristics). In our context, this means that any increase in member cohesion

might be product of better whipping or agenda control, or a product of MPs sorting themselves

into parties with policies more suited to their underlying preferences.

5

For foregoing scholars, solving this problem was a tall order since it implicitly required records of

individual MP behavior, over a long time period. That is because the above debate—‘inducements’

vs ‘replacement’ effects—can only be resolved by knowing whether MPs became more disciplined

over their career in the House, and whether or not that evolution outweighs (in a statistical sense)

the effect of new members entering the Commons who were ‘already’ more cohesive in type. Put

more simply, we need panel data and previous (sometimes Herculean) efforts have either limited

the scope to one or two parliaments with records at the MP-level (e.g. Adyelotte, 1954; McLean,

2001; Schonhardt-Bailey, 2003), or have looked at a longer time period in aggregate terms without

attention to unit (MP) level variation (e.g. Lowell, 1902; Berrington, 1968; Cox, 1987). There are

exceptions Cromwell (1985), but the coverage of both bills and years is spotty and truncated. Quite

apart from issues of data availability, there are methodological difficulties in comparing across years

and parliaments. In particular, most scholars (e.g. Cox, 1987) have typically used means of cohe-

sion, or have looked at simple counts Lowell (1902); Berrington (1968) of certain types of division.

Such measures ignore variation in the tails of the distributions of MPs they describe: variations in

the most loyal and least loyal members are not modeled in any explicit way, with obvious problems

for determining questions of cohesion over time.

As we explain in the next two sections we improve on both areas, by bringing to bear the en-

tire universe of MP decisions on roll calls for the period 1836–1910 and new methods to study

them. This allows us to model cohesion ‘within’ MP careers. Thus, if we believe that latent

preferences are essentially fixed, and we can show that cohesion increases are not the product of

replacement effects, we can go some way to showing that an inducements story—in which whipping

and other institutions became more powerful motivation tools—is more convincing. More formally,

our intention is to

1. examine the association between being a newcomer to the House of Commons and voting,

in terms of average cohesion by member. If the replacement story has merit, fresh entrants

ought to, ceteris paribus, be more disciplined than the mass of colleagues they join who are

already incumbents;

6

2. examine the effect of serving in one’s last parliament as a predictor of cohesion. If the

replacement story embodies the truth, then retirees (those leaving the chamber for the last

time) ought to be, on average, less disciplined controlling for their seniority, age and other

factors;

3. examine the association between seniority and discipline for individual MPs over time. Here,

the ‘inducements’ theory predicts that whatever their ‘type’, MPs will become more disci-

plined (on average) over time as new internal organizational incentives begin to bite. That

is, controlling ceteris paribus MPs will respond to changing circumstances by obeying their

whip more often.

3 Data

The current project requires roll call votes of MPs in the House of Commons, or ‘divisions’ as they

are known for this particular legislature. As noted, several authors have collected roll calls for the

19th Century prior to our efforts, but we are much more comprehensive in both depth and breadth.

In particular, we gathered every individual roll call vote for the entire period 1836–1910.

Our raw data was supplied by the History of Parliament Trust, and in that original form in-

cludes scanned pages of division lists which were fed through optical character recognition to form

text documents of names, listed under the relevant ‘Aye’ or ‘No’ column in Hansard. There are

some 20262 of these roll calls. Unfortunately, many names are ambiguous: in particular, names

are misspelled or presented as non-standard contractions in multiple different ways. While much

disambiguation could be performed via automatic, computer-based methods, we required the use

of a large number of research assistant hours to manually assign the appropriate (unique) identities

to the names given. The identities of the MPs in the roll call sheets were matched up to names

in our broader database, that lists every MP serving during this period, and some key background

covariates pertaining to that MP (date of birth, date of death, dates of service, ministerial rank,

constituency name, election results, party etc). The covariates themselves come for various stan-

7

dard sources, such as Craig (1989), Craig (1974), Walker (1978), Cook and Keith (1975) and Butler

and Butler (1994). In this way, each vote in each division list becomes part of a record on a given

individual. By the end of the cleaning process, both automatic and manual, missingness had been

reduced to around one percent of votes; that is, we were uncertain who the MP cast was in around

one percent of the 4, 791, 282 vote decisions in our data.

For reasons that will become clear below, we required the identity of the head (‘chief’) whips

for the two major parties (Liberals and Conservatives for this period) at any given time. We found

annual listings of their identities in Cook and Keith (1975) and Butler and Butler (1994) but the

precise resignation and promotion dates for these officers is unknown. While it is generally the

case that government Chief Whips are made parliamentary secretary to the Treasury, obtaining

dates for this office proved no easier. And, in any case, there is no such aliased information for

Chief Whips of non-government parties at this time. As a result, we occasionally found months of

roll calls within years in which the particular whip cited in aggregate records did not appear to be

serving as an MP (and we therefore assume he was not in fact whip during this period). When

that occurred, we simply assumed that the ‘next’ whip of the party already held that position

at the time of the roll call. Empirically, this may be correct or not (there is no way of knowing

with current data), but it is unlikely to cause problems for our measure below since, presumably,

those who become Chief Whip vote in a way congruent to that office immediately preceding their

ascension to it.

3.1 Measuring Discipline

Measuring the cohesiveness of voting in British parliamentary parties is a long-standing method-

ological problem (see,e.g., MacRae, 1970, for an overview). Most scholars have suggested measures

which use aggregate information on parties. That is, they calculate the discipline of one party on

a given roll call as a whole and compare it to another. Here, the Rice Index is the most commonly

used such approach, defining the discipline on division i as|ayesi−noesi|ayesi+noesi (see Desposato, 2003, 2005;

Hix, Noury and Roland, 2005, also). The numerator, and thus the index as a whole, is increasing

8

in the difference between the number of MPs voting ‘aye’ and ‘no’, reaching a minimum of zero

in the event that opinion is exactly evenly split. By contrast, other measures explicitly contrast

multiple parties on a given vote; thus, for example, Lowell (1902) considers the proportion of roll

calls in which 90 percent of one party’s MPs vote contrary to 90 percent of another’s.

Our goal in the current paper is to shed light on the way that individuals behave with respect

to their parties, and thus we will not be directly concerned with the aggregate measures above.

Instead, we begin from the premise that an MP votes in inline with his party when he votes inline

with his party Chief Whip, who we assume receives orders directly (or as if directly) from his party

leader. It could conceivably be the case that a particular roll call was not whipped, yet the Chief

Whip chose to vote ‘freely’. In that case, voting with the Chief Whip overstates the degree of

discipline pertaining to a specific MP. But such occurrences are likely to be extremely rare, since

Chief Whips are agents of leadership throughout this period and unlikely to waste time on more

frivolous matters that come up for roll call votes. Alternatively, it may be the case that, on a

particular whipped roll call, the Chief Whip fails to vote himself for some idiosyncratic reason, yet

gives instructions to the MPs (which we cannot observe). We assume such events are rare, or are

missing at random from our data.

To clarify how our measure works, consider the following example. Let w be the whip voting

vector. That is, the vector of decisions on all the roll calls in a given parliament. Any member,

including the whips, may vote one of three ways: ‘aye’ (coded as 1), ‘no’ (coded as 0), or not voting

(coded as NA). In what follows, suppose there are a total of five roll calls and that the whips’

voting profile is

w = [1, 0,NA︸ ︷︷ ︸whip1

, 1, 1︸︷︷︸whip2

].

9

Here, there were two Chief Whips in office during the parliament in question,1 and thus the vector

is implicitly divided into two parts. Suppose a Member of Parliament i’s voting vector is

mi = [1,NA, 0, 1, 0].

Since MP i only votes in four divisions, the denominator for any agreement scores is four. Here, i

votes in the same way as the whip on two of the four roll calls on which he attends, and thus i’s

agreement score is 24 . Notice that missingness in the MP’s vote vector does not count against the

MP’s disciplinary record. Similarly, when the whips’ votes are missing, that roll call is removed

from consideration in terms of agreement. The distribution of all such MP’s agreement scores is de-

noted as the “discipline distribution” below. Note that this aggregate distribution must be bounded

between zero and one. Importantly, while the whip vector is party-specific, the aggregate discipline

distribution merges the parties. That is, the aggregate distribution has as its ith component the

agreement score between the i MP and his party-specific whip, though that party can be different

from MP to MP. This means that a Tory MP with the same discipline score as a Liberal MP in

a given parliament may have voted in exactly the same way in practice as the Liberal (‘ayes’ and

‘noes’ in the same places), or in the exactly converse way, or something in between. What matters

is how often he follows his whip, and whips in different parties may in some circumstances agree

(in voting terms) on specific issues. Finally, note that we do not require both party whips to vote

in a given division for it to be included in our analysis—so long as one enters a preference, that

division enters the data for his party.

As our comments above imply, we restrict the analysis that follows to the two major governing

parties at this time: the Conservatives and Liberals. While we have information on MPs from

other parties, we typically do not know their whipping arrangements, and the process is especially

complicated for some ‘break away’ factions like the Liberal Unionists who spend some time in the

19th Century voting with the Conservative whip. Thus our restriction avoids such ambiguities.

1Note that our scheme only allows one whip as the Chief Whip at any particular time, so there is no danger ofambiguous cases should two people who held the office (eventually) happen to disagree on a specific roll call.

10

●

●●

●

●●

●

●

●

●

●

●

●

●

●

●

●●

●

●

●●●

●

●

●

●

●

●

●

●

●

●

●

●●

●●●

●

●

●

●

●●●

●

●●

●●

●

●●

●

●

●

●●

●●

●●

●

●

●

●●

●

●

●●

●●●●

●

●●

●

●●

●●●

●●

●

●

●

●

●●●

●●●●

●

●

●

●●●

●

●

●

●

●●●

●

●

●

●

●●

●

●●

●

●

●●●

●

●●●●

●

●

●●

●

●

●●

●●●●●●

●

●●

●

●●●●

●

●

●

●●●

●

●

●

●

●

●●

●

●

●●●●

●

●

●

●

●●●

●

●

●●●

●

●

●

●

●●●

●

●

●

●●

●

●

●●

●

●

●

●

●

●

●

●●

●

●

●

●●

●

●

●

●

●

●

●

●●

●●

●●

●●●

●

●●

●●●●

●

●

●

●

●

●●●

●

●

●

●

●

●●●●●

●

●●●

●

●

●●

●

●●

●

●●

●●●●●

●

●●

●

●●

●

●●●●●●●●●●●●

●

●

●●

●●●

●●

●●

●

●

●

●

●●●

●●●●●

●●●

●●

●●●●●●●●●●●●●●●●●●

●

●●●

●

●●●●●●

●

●●●●

●

●

●

●

●●●●●●

●

●●●●●●

●

●●●●●

●

●

●

●

●●●

●●

●●●●●●●●●

●

●

●

●●●●●●

●●●●●

●

●●

●

●●●

●

●

●●●

●

●●

1835

1837

1841

1847

1852

1857

1859

1865

1868

1874

1880

1885

1886

1892

1895

1900

1906

0.0

0.2

0.4

0.6

0.8

1.0

Figure 1: Aggregate discipline in the House of Commons over time. Measure bounded between 0and 1, with 1 being most disciplined.

In Figure 1, we report boxplots representing the distribution of the scores introduced above, over

time. The thick black bars represent medians, and it is clear that while median behavior becomes

less disciplined in the 1840s, it approaches 1 thereafter, reaching its zenith at the end of the data.

The grey points represent outliers who are least disciplined: MPs whose record puts them into the

tail of their parliament’s distribution. While the pattern observed—more disciplined MPs packed

more tightly around a higher median—is relatively intuitive, we wish to be more formal about the

changes to the distributions. In the next section, we do just that.

11

4 Methods

The essence of our task is to compare the distribution of MPs in one parliament, in terms of their

obedience and cohesiveness, to the distribution of MPs in another. Here, we suggest a ‘relative

distribution’ approach and we follow the presentation of Handcock and Morris (1998), albeit with

a slight change in notation to fit the intuition of our substantive example.

4.1 Relative Distribution Approach

Denote the ‘proportion of whip agreement’ for the ith MP as yi. The goal is to compare the MPs

from a parliament in the previous period, t−1, with one in the next period, t. For the earlier parlia-

ment, denote the cumulative distribution function for these MPs as Ft−1(y). Let Yt−1 be a random

sample of whip agreement scores from it. The CDF for the next parliament is Ft(y), and Yt is a ran-

dom sample from it. First assuming that Ft−1 and Ft are continuous with common support (here,

[0, 1]), we define the grade transformation of Yt−1 to Yt as R = Ft−1(Yt). Note then, that the grade

transformation R is the result of plugging Yt into the earlier CDF Ft−1: for any given value of Yt

(say, 0.81), we get back the proportion of observations in t−1 that are at or below this value (that, is

the proportion of MPs in the earlier parliament that agree with their whips 0.81 or less of the time).

The next goal is to express the CDF and probability density function (PDF) of R itself. In

particular,

G(r) = Ft(F−1t−1(r)) 0 ≤ r ≤ 1, (1)

where r is the realization of R, known as the relative data, and F−1t−1(r) is the inverse CDF for t−1.

In words, r can be thought of as a percentile rank that the ith MP in parliament t would have had

in the previous parliament. By taking F−1t−1(r), we obtain the actual values of whip agreement at

or below that value of r in the earlier parliament. Finally, by taking Ft(F−1t−1(r)), we obtain the

proportion of the MPs in parliament t lying at or below the rth quartile of MPs in the previous

12

0.0 0.5 1.0

01

23

45

current parliament

previous parliament

Figure 2: Example PDFs for two parliaments: current parliament (yellow, front of figure) hashigher mean and variance than previous parliament (gray, back of figure)..

parliament. The PDF of R is

g(r) =f(Ft−1(r))

ft−1(F−1t−1(r))

0 ≤ r ≤ 1. (2)

Note that the PDF is simply a density ratio: the proportion of MPs in the current parliament

relative to the proportion of MPs in the previous parliament, at a given level of whip agreement.

To clarify matters, consider the following (simulated) data example. Suppose that MPs from

the previous parliament have whip agreement scores that are distributed normal, with mean 0.5,

standard deviation 0.08; MPs from the current parliament are distributed with a standard deviation

of .2, and their mean is slightly higher (more disciplined) at 0.6. The sample distributions for these

MPs are given in Figure 2. Following the steps above, it is straightforward (in this parametric case)

to obtain the G(r) and g(r), CDF and PDF of the relative data, respectively. In Figure 3 we plot

13

0.0 0.2 0.4 0.6 0.8 1.0

0.2

0.4

0.6

0.8

1.0

Proportion of previous parliament

Pro

port

ion

of c

urre

nt p

arlia

men

t, G

(r)

Figure 3: Example CDF for relative data from the two parliaments shown in Figure 2. The dottedline demonstrates that 0.6 (60%) of the MPs in the previous parliament were at or below the 40thpercentile in the current parliament, in terms of their whip agreement scores.

the CDF of the relative data: notice that the x-axis is the proportion of the previous parliament,

while the y-axis is the proportion of the current parliament. Thus, the unbroken CDF line is the

relationship between them. Notice, for example, that only (approximately) 36% of MPs in the

current parliament have a whip agreement score at or below the 60th percentile of the previous

parliament. Similarly, we can investigate the relative PDF, and Figure 4 presents this based on

our simulated data. When the PDF is above one, there is more density, for the given relative data

range, in the current parliament. The ‘U’-shape here makes sense: looking at Figure 2, we see

that the current parliament has tails well above and below those of the previous parliament. As a

result, the relative PDF in Figure 4 is well above one in its tails: in the current parliament, there

are approximately two or three times as many MPs with scores in the 0th to 4th percentile of the

previous parliament (0–0.4 as a proportion) as in the previous parliament itself; similarly in the

current parliament, there are two to three times as many MPs with scores from the 95th to 100th

percentile of the previous parliament than in the previous parliament itself.

14

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.5

1.0

1.5

2.0

2.5

3.0

Proportion of previous parliament

Rel

ativ

e D

ensi

ty

Figure 4: Example PDF for relative data from the two parliaments shown in Figure 2. When thePDF is above one, there is more density in the current parliament for a given percentile in theprevious parliament, when it’s below one there is more density in the previous parliament.

Relative distributions capture the differences between data in an intuitive way, but there are

other reasons to employ them. In particular, they are non-parametric, meaning that one does

not need to commit to, for example, looking at mean and variance shifts under the very dubious

assumption that whip agreement scores are normal. Indeed, there are no such assumptions about

the relationship between the current and previous parliaments. Second, and we shall now see, they

can be studied with summary measures that are robust to outliers—and important characteristic

in the context of voting in parliament.

4.2 Decomposing the relative distribution

The relative distribution relates data from the current parliament back to that of a previous par-

liament. While visual summaries like the above are helpful, there is obvious interest in having

a more formal way of measuring how the distributions in question differ. Handcock and Morris

(1998) suggest two measures, that pertain to location and shape changes. Location changes refer

to notions of a ‘shifted’ distribution, while shape changes refer to more general alterations to the

spread, size of the tails and so on. Importantly, the totality of the changes from the previous to

15

the current parliament are completely captured by these two effects, meaning that we can talk

meaningfully about the contribution of the location and shape component. In practice, the analyst

can chose a location measure to match the substantive problem at hand, and here we consider the

median.

The specific summary measure suggested for comparing the contribution of the two effects is based

on the Kullback-Leibler divergence between the two distributions in question and details can be

found in Handcock and Morris (1998) along with information on the computation of uncertainty

estimates. Intuitively, one can think of the techniques used in the following way: first, suppose

we take the MPs of the previous parliament (Yt−1), and shift their distribution of whip agreement

scores such that they have the same median as the current parliament, but do nothing to the shape

of the distribution which remains whatever it was for the previous parliament. Call that set of

scores Yshift. Comparing Yt−1 to Yshift allows us to isolate the effect of the location change. Com-

paring Yt, the scores of the current parliament, to those of Yshift allows us to isolate the effect of

the shape change because, by construction, Yt and Yshift have the same location so any difference

between them arises from a change to some other moment of the distribution.

The other summary measure Handcock and Morris (1998) suggest pertains to the ‘polarization’

of the distribution over time. In our context, their ‘polarization index’ refers to the extent to which

there is more mass in the tails of the current parliament distribution than in the previous parlia-

ment. Thus, when the index is rising, we are in a situation where MPs are increasingly different

from one another: spread over the support of the whip scores increasingly thinly. By contrast,

when the index is falling, MPs are increasingly gathering around the central tendency, whatever

that location measure might be. Such approaches are implemented in the R statistical environment

by Handcock and Aldrich (2002), and we use their software in what follows.

16

5 Results

Before discussing our results, we note that one drawback of using relative distribution methods is

that they tend to require large amounts of data, relative to parametric alternatives. Second, they

are sensitive to ‘heaping’ (Handcock and Morris, 1998, 89), which refers to the tendency of social

scientific data to ‘pile up’ in certain places on the continuous spectrum. Here, as time goes on, more

and more MPs appear at, or close to, a whip agreement score of 1 (that is, full agreement). As a

result, the accuracy of the entropy calculations—a measure of the distance between two distributions

that can be decomposed into location and shape components—tends to suffer. In particular, for

some of our comparisons, the (overall) entropy or contribution of some of the components is reported

as negative (logically impossible, given the way the measure is conceived). In Table 1 we report

the actual figures as estimated, despite the fact that the point estimates are mis-signed in some

cases. Notice though, that even in such cases, we can compare the absolute size of the component

contributions, to get a sense of their relative importance for the overall (‘Total Entropy’) change.

For every comparison, bar the final parliament at the end of the data, the shape shift contributes

more to the changing distribution than the median shift. That is, to the extent that we see changes

over time in the way that MPs agree with their whips, it is not the average that varies so much

as other distributional characteristics. For clarity, here and below, we refer to the parliament that

began in 1835 as the ‘1835 parliament’, the one that began in 1837 as the ‘1837 parliament’ and so

on.

In Figure 5, we provide a barplot of the same information as in Table 1 except that we impose

the rationality constraint that all negative values (of location contribution) are in fact zero, and

then treat the sum of the location and shape contributions as the total entropy. The point is clear

from the plot: with one exception, at the end of the data, all the changes to the parliamentary

distributions of discipline come primarily through shape changes, rather than location shifts.

17

Previous Current Total location shapeparl parl entropy contribution contribution

1835 1837 0.06 -0.03 0.061837 1841 0.10 -0.11 0.041841 1847 0.16 0.02 0.151847 1852 0.15 -0.92 0.201852 1857 0.04 0.01 0.041857 1859 0.07 -0.03 0.111859 1865 0.09 0.04 0.421865 1868 0.02 -0.01 0.041868 1874 0.15 0.09 0.101874 1880 0.13 -0.08 0.221880 1885 0.47 -0.65 0.591885 1886 0.46 -0.19 0.651886 1892 0.32 0.04 0.301892 1895 0.23 0.04 0.151895 1900 0.20 -0.09 0.301900 1906 0.23 0.12 0.10

Table 1: Relative contribution of median and shape to changing distribution of whip agreementscores over time. Note that the negative values arise from the data requirements of relative distri-bution methods not being fully met: note though, that the relative contribution of the componentscan be still be compared in absolute terms.

1835

1837

1841

1847

1852

1857

1859

1865

1868

1874

1880

1885

1886

1892

1895

1900

locationshape

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

Figure 5: Barplot showing relative contribution of location and shape components in relative dis-tributions for parliaments in the data.

18

5.1 The ‘concentration’ of discipline

As noted above, we can also study the ‘polarization’ of a distribution: essentially a measure of

what proportion of the distribution’s mass is in the center relative to the tails. Handcock and

Morris (1998) suggest three such measures: the median relative polarization index, which varies

between −1 and 1, wherein a positive value implies more mass in the tails, and a negative value

implies more concentration around the median. Similarly, they suggest indices pertaining to the

upper and lower tails with range and interpretations in the same way as the median index. We

investigated the change in all of these indices for our data, which we arrange is a slightly different

way to the above discussion. In particular, we now treat each parliament as the ‘current’ one, while

all those before it are aggregated into the ‘previous’ parliament. The idea is to ask the question

“how does the current parliament compare, in distribution terms, to all those coming before it?”.

Figure 6 presents the results. There, circular points represent the median index, triangles represent

the lower tail index, and squares represent the upper tail index. When filled, the points imply that

the index is statistically significant (meaning that any observed polarization of the distribution—

i.e. deviation in shape from the previous parliament—is sufficiently unlikely to have occurred by

chance). We note that these indices generally move together, but as the solid lowess line makes

clear, they are decreasing over time. In practice, this means that the discipline distribution for the

MPs is becoming increasingly concentrated around the center of that distribution.

An important observation is that due to the nature of the densities in question, the lower tail

polarization is much more informative than the equivalent for the upper tail. Because the center

of the distribution is already so close to its maximum, the upper tail index can mislead: as the

median approaches one (which it does, empirically, by the 1860s), the upper tail cannot get more

populated since essentially all the weight of the distribution is to the left of the center. Thus, it

is the lower index which we ought to pay most attention to, and this is quite clearly declining

over time as rebelliousness, measured by deviation from the whip, virtually disappears from the

discipline distribution.

19

●

●

●

●●

●

●● ●

●

●

●

●

●

●●

1840 1850 1860 1870 1880 1890 1900

−1.

0−

0.5

0.0

0.5

●

lower tailmedianupper tail

Figure 6: Share of the discipline distribution change in the tails, over time. Comparison is currentparliament to all previous ones at that point. Circles represent median index, triangles the lowertail index and squares the upper tail index. Solid symbols imply effects with p < 0.01, non-filledsymbols imply otherwise. Note the solid lowess line composed of all indices, decreasing over timeand implying that the distribution is becoming more concentrated around its median.

20

Figure 6 and Table 1 tell a consistent story: as the 19th Century progressed, the distribution

of the discipline in the House of Commons became increasingly concentrated around its average;

however, the average itself (measured as the median) did not change greatly. Already high in the

1830s, most of the change in MP behavior came in the bottom tails. Put otherwise: for whatever

reason, over the period we study the norm was almost always to vote as the whips did; what

changed was the proportion of MPs who did not obey this norm. And such behavior become more

scarce as the century progressed.

5.2 Decomposing the Decomposition

Recall that if the ‘replacement’ story explains the decreased number of ‘rebels’ in the left tail, we

ought to see (1) ‘new’ MPs that are on average more disciplined than the parties they were joining

and (2) ‘leavers’—that is MPs leaving the chamber—that are on average less disciplined than the

parties they were exiting from. To test for the first mechanism, we regressed the discipline scores for

MPs on their binary status as newcomers to parliament (or not); we did this for every parliament.

The results can be seen in Figure 7, where we plot the regression coefficients, for Liberals and Con-

servatives over time, with their respective 95 % confidence intervals. When the confidence interval

is everywhere above the zero line, being a newcomer is associated with being more disciplined (on

average) than having previously served. Without exception, new entrants to the House of Com-

mons are not more disciplined. In the rare cases where the estimates are statistically differentiable

from zero, those incoming cohorts (Liberals in the mid-1830s, 1840s and mid-1880s) are actually

less disciplined than their colleagues.

Second, we investigate the notion that ‘leavers’ are less disciplined on average. In particular,

for every parliament we treat service in the subsequent parliament as the dependent variable (thus

y = 1 if an MP continued to serve, y = 0 if an MP exited). On the right hand side, we use the MP’s

discipline score along with their seniority (years served) and their age as controls. We also include

a parliament fixed effect. The results—in terms of the coefficient estimates on discipline score—are

reported in Figure 8. With some exceptions in the 1840s of inconsistent directions, there is scant

21

Parliament end points

Coe

ffici

ent o

n 'n

ew'

1835 1841 1847 1852 1857 1865 1874 1880 1885 1892 1900 1906 1910

−0.

10−

0.05

0.00

0.05

●

●

● ●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

ConLib

Figure 7: Regression coefficients: discipline score regressed on binary new-comer to Commons sta-tus, point estimates and 95% confidence intervals. Liberals are [brown] dashed lines; Conservativesare unbroken [blue] lines. When the confidence interval is everywhere above the zero line, being anewcomer is associated with being more disciplined (on average) than having previously served.

evidence of an effect: that is, the 95% confidence intervals typically overlap with zero, and we have

little reason to believe that those leaving the chamber were more likely to be ‘rebels’ (low discipline

types) than others, on average.

With the above in mind, we are logically left with one mechanism: if entrants were not more

disciplined, and leavers were not less disciplined, essentially all the increase in whip agreement

must come ‘within’ MP careers. That is, the average MP must become more disciplined during the

time he served in parliament, implying that he leaves parliament more disciplined than he begun it.

To test this supposition, we treat the discipline data as an unbalanced panel: that is, a time-series

(by parliament) of discipline observations on MP units. Because MPs might have different latent

propensities to rebel or be loyal, we utilize fixed effects. In a first model, we regress discipline on

seniority (number of years served), along with the varying intercept term. In a second model, we

interact seniority with a dummy for any observation prior to 1852. Our idea here is to control for

the ‘special’ period in the 1840s wherein discipline was generally lower as noted earlier. Our results

are as Table 2. In the first model, seniority is statistically significant: with every year of service

that passes, MPs become more loyal. In the second model, we see that the implied coefficient on

the seniority for the period after 1852 increases relative to the period before that date: it is 0.0033

22

Parliament end points

Coe

ffici

ent o

n 'd

isci

plin

e'

1835 1841 1847 1852 1857 1865 1874 1880 1885 1892 1900 1910

−3

−2

−1

01

23

●

●

●

●●

●

●

●●

● ●

●

●

●

● ●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

ConLib

Figure 8: Coefficient estimates, with 95% confidence intervals, on discipline as a predictor of‘survivor’ status to next parliament.

for the second half of the 19th Century, and 0.0033 − 0.0036 = 0.0003 for the first half. This

implies we do indeed have a period effect, but even allowing for it, as MPs serve over the course

of the century, they tend to be more loyal. One way to think about the importance of service as a

predictor is to note between 1852 and 1906, mean discipline increased from around 0.80 to around

0.97, a gain of 0.17. But that is almost exactly equal to the implied gain from our model for an MP

serving (hypothetically) the entire period under study: that is, β̂post-52 × (1906 − 1852) = 0.18.

Put otherwise, our model here fully explains the increase in discipline for the second half of the

19th Century with no recourse to, and no evidence for, replacement effects: it is simply that MPs

became more loyal as a secular trend throughout the Victorian era.

5.3 Summary of Results

The foregoing sections showed that

1. average cohesiveness—measured as the proportion of times an MP voted with his whip—was

always relatively high, with a baseline of around 0.9 in 1830s. It dipped in the 1840s, and

increased to present day levels thereafter;

2. almost all the aggregate increase in cohesion is a product of the left tail of discipline scores

receding into the main mass of the distribution: that is, rebels became fewer and further

23

Model 1 Model 2(interaction)

seniority 0.0019∗∗∗ 0.0033∗∗∗

[0.0001] [0.0002]

before 1852 0.0760∗∗∗

[0.0054]

seniority× before 1852 −0.0036∗∗∗

[0.0004]

adj-R2 0.0174 0.0355

Table 2: Regression of discipline on seniority (and MP fixed effects). Model 1: includes onlyseniority. Model 2: includes seniority and interaction of seniority with parliaments before 1852.β̂∗∗∗ implies coefficient significant at the 0.01 level, standard errors in parentheses.

between over time;

3. there is no evidence of a ‘replacement’ effect: incoming MPs, throughout the century, were

not more disciplined nor were retiring ones less so;

4. there is, however, strong evidence of an ‘inducements’ effect in which MPs became more

cohesive with respect to their own party over time, within their own parliamentary careers.

6 Discussion

Legislative party cohesion is important for the functioning of Westminster democracies. Without

it, the link between government policy-making and voter preferences is broken. Understanding how

this institutional arrangement came about has therefore been of critical importance to scholars of

history and political science (e.g Cox, 1987; Jenkins, 1996; Adelman, 1997). Here, we undertook

a massive data collection task to shed light on the central question of why members became more

disciplined. We gathered over 20, 000 roll calls from the period 1836–1910, and we matched them

to individual MP records so that intra-career changes in cohesion could be calculated for the first

time. We compared these ‘inducement’ effects—deriving as we think they do from more efficient

institutional mechanism for disciplining and controlling backbenchers—to more sociological ‘re-

placement’ effects, by which new personnel with new characteristics might drive the patterns we

24

saw. In practice, we were able to reject essentially all claims that the latter school of thought might

offer. We showed that what drove cohesion was the ‘folding in’ of the rebellious left tail of the

cohesion distribution, and that this general pattern was consistent over time and was not a product

of replacement dynamics.

Our finding has several broader implications. First, whatever parties did to bring their back-

benchers into line—be it better whipping, more palatable agenda setting, increasingly efficient

incentives etc—it worked. There is clear evidence of learning, and the notion that actors within

institutions both influence, and are influenced by, their surroundings is apparent from our work.

In that sense, it joins a swathe of literature on ‘rational choice institutionalism’ beginning at least

with Riker (1980) (see Shepsle, 2006, for an overview). Second, to the extent that we care about

the recruitment of different types of individuals to Westminster parliaments and use measures such

as “all women short-lists” to achieve this end (see Norris and Lovenduski, 1995; Lovenduski, 2005,

for discussion), it is not obvious that this will change the way ‘business as usual’ occurs in these

places. That is, we find that this institution of governance is remarkably robust to new entrants

and thus, we suspect, any different operating principles they might have (see also Cowley, 2002,

on this point). Third, whatever the electoral consequences of the Second Reform Act (1867) (see

Smith, 1967; Feuchtwanger, 1968; Jenkins, 1996; Berlinski and Dewan, 2011), the legislative effects

of this surge in the franchise appear to have been minimal as per Figure 1, with parties already on

the way to modern practice by the 1850s. This is in line with suggestions from others Cox (1987),

albeit we have more comprehensive data to make the point.

What we have not uncovered here is the precise micro-mechanism that compelled backbench troops

to fall into line. There are certainly numerous plausible theories, ranging from a desire to obtain

ministerial portfolios (Cox, 1987), to socialization (Rush and Giddings, 2011), to partisanship

(Jenkins, 1996), or some combination of the above (Kam, 2009b). Absent more detailed survey

work on (long dead) MPs, fleshing out the plausibility of these ideas is difficult. Nonetheless, op-

erationalizing some of these accounts in the data we have may not be impossible for the later part

25

of our period. One option is to use parliamentary speech to get a sense of motivations and trends

over time, and compare those records with division behavior for individual MPs. We leave such

efforts for future work.

26

References

Adelman, Paul. 1997. Gladstone, Disraeli and Later Victorian Politics. New York: Longman.

Adyelotte, W.O. 1954. “The House of Commons in the 1840s.” History XXXIX:249–262.

Aydelotte, William O. 1963. “Voting Patterns in the British House of Commons in the 1840s.”Comparative Studies in Society and History 5(2):134–163.

Bagehot, Walter. 1873/2011. The English Constitution (2nd Edn). Accessed June 6, 2011.:http://socserv.mcmaster.ca/ econ/ugcm/3ll3/bagehot/constitution.pdf.

Benedetto, Giacomo and Simon Hix. 2007. “The Rejected, the Ejected and the Dejected: ExplainingGovernment Rebels in the 2001-5 British House of Commons.” Comparative Political Studies40(7):755–781.

Berlinski, Samuel and Torun Dewan. 2011. “The Political Consequences of Franchise Extension:Evidence from the Second Reform Act.” Quarterly Journal of Political Science 6(34):329–376.

Berrington, Hugh. 1968. “Partisanship and Dissidence in the Nineteenth Century House of Com-mons.” Parliamentary Affairs XXI:338–74.

Birch, A. H. 1964. Revresentative and Responsible Government. London: George Allen & Unwin.

Bogdanor, Vernon and David Butler. 1983. “Democracy and elections: Electoral systems and theirpolitical consequences.”.

Butler, David and Gareth Butler. 1994. Twentieth Century British Historical Facts. 8th ed. London:MacMillan Press.

Campbell, Colin and Graham Wilson. 1995. The End of Whitehall: Death of a Paradigm. Hoboken,NJ: Wiley-Blackwell.

Cook, C. and B. Keith. 1975. British Historical Facts 1830-1900. London: Palgrave-MacMillan.

Cowley, Philip. 2002. Revolts and Rebellions: Parliamentary Voting under Blair. London: Politicos.

Cowley, Philip. 2005. The Rebels: How Blair Mislaid His Majority. London: Politico’s PublishingLtd.

Cox, Gary. 1987. The Efficient Secret: The Cabinet and the development of Political Parties inVictorian England. Cambridge: Cambridge University Press.

Cox, Gary and James Ingram. 1992. “Suffrage Expansion and Legislative Behavior in Nineteenth-Century Britain.” Social Science History 16(4):539–560.

Craig, Frederick. 1974. British Parliamentary Election Results, 1885–1918. London: MacMillanPress.

Craig, Frederick. 1989. British Parliamentary Election Results, 1832–1885. London: MacmillanPress.

27

Cromwell, Valerie. 1982. “Mapping the Political World of 1861: A Multidimensional Analysis ofHouse of Commons’ Division Lists.” Legislative Studies Quarterly 7(2):281–297.

Cromwell, Valerie. 1985. “Computer Analysis of House of Commons Division Lists 1861-1936: areport on current research.” Parliaments, Estates and representation 5(1):33–36.

Desposato, Scott. 2003. “Comparing Group and Subgroup Cohesion Scores: A NonparametricMethod with an Application to Brazil.” Political Analysis (11):275–288.

Desposato, Scott. 2005. “Correcting for Bias in Rice Cohesion Scores.” British Journal of PoliticalScience 35(4):731–744.

Diermeier, Daniel and Timothy Federsen. 1998. “Cohesion in Legislatures and the Vote of CondenceProcedure.” American Political Science Review 92:611–622.

Economist. 2012. “A damn good whipping.” Bagehot, Economist Nov 3.

Feuchtwanger, Edgar. 1968. Disraeli, Democracy, and the Tory Party. Oxford: Clarendon Press.

Gamble, Andrew. 1990. “Theories of British Politics.” Political Studies 38:404–420.

Handcock, Mark and Eric M Aldrich. 2002. “Applying relative distribution methods in R.” Uni-versity of Washington, Center for Statistics and the Social Sciences Working Papers (27).

Handcock, Mark and Martina Morris. 1998. “Relative Distribution Methods.” Sociological Method-ology 28:5397.

Handcock, Mark and Martina Morris. 1999. Relative Distribution Methods in the Social Sciences.New York: Springer.

Hanham, Harold. 1978. Elections and Party Management. Harvester Press.

Hix, Simon, Abdul Noury and Gerard Roland. 2005. “Power to the Parties: Cohesion and Compe-tition in the European Parliament, 1979-2001.” British Journal of Political Science 39:209–234.

Jenkins, Terence. 1996. Parliament, party and politics and Victorian Britain. New York: Manch-ester University Press.

Kam, Christopher. 2009a. “Partisanship, Enfranchisement, and the Political Econ-omy of Electioneering in the United Kingdom, 1826-1906.” Working Paperwww2.lse.ac.uk/government/research/resgroups/PSPE/pdf/Kam.pdf.

Kam, Christopher. 2009b. Party Discipline and Parliamentary Politics. Cambridge: CambridgeUniversity Press.

Krehbiel, Keith. 1993. “Where’s the party?” British Journal of Political Science 23:235–66.

Lijphart, Arend. 1999. Patterns of Democracy: Government Forms and Performance in Thirty-SixCountries. New Haven, CT: Yale University Press.

Lovenduski, Joni. 2005. Feminizing Politics. Cambridge: Polity Press.

28

Lowell, James. 1902. “The Influence of Party on Legislation in England and America.” AnnualReport for the American Historical Association for the year 1901 i.

MacRae, Duncan. 1970. Issues and Parties in Legislative Voting: Methods of Statistical Analysis.New York: Harper & Row.

McLean, Iain. 2001. Rational Choice and British Politics. Oxford: Oxford University Press.

Norris, Pippa and Joni Lovenduski. 1995. Political Recruitment: Gender, Race and Class in theBritish Parliament. Cambridge: Cambridge University Press.

Norton, Philip. 1978. Conservative Dissidents: Dissent within the Parliamentary ConservativeParty, 1970–74. London: Temple-Smith.

Powell, Bingham. 2000. Elections as Instruments of Democracy: Majoritarian and ProportionalVisions. New Haven: Yale University Press.

Rhodes, R.A.W., John Wanna and Patrick Weller. 2009. Comparing Westminster. Oxford: OxfordUniversity Press.

Rhodes, R.A.W. and Patrick Weller. 2005. Westminster Legacies: Democracy and ResponsibleGovernment in Asia and the Pacific. University of New South Wales: University of New SouthWales Press chapter Westminster Transplanted and Westminster Implanted: Exploring PoliticalChange.

Richards, David and Martin Smith. 2002. Governance and Public Policy in the United Kingdom.Oxford: Oxford University Press.

Riker, William. 1980. “Implications from the Disequilibrium of Majority Rule for the Study ofInstitutions.” American Political Science Review 74:432–46.

Rush, Michael. 2001. The Role of the Member of Parliament Since 1868: From Gentlemen toPlayers. Oxford: Oxford University Press.

Rush, Michael and Philip Giddings. 2011. Parliamentary Socialisation. New York: Palgrave-MacMillan.

Schonhardt-Bailey, Cheryl. 2003. “Ideology, Party and Interests in the British Parliament of 1841-1847.” British Journal of Political Science 33:581–605.

Shepsle, Kenneth. 2006. Oxford Handbook of Political Institutions. Oxford: Oxford UniversityPress chapter Rational Choice Institutionalism.

Smith, Paul. 1967. Disraelian Conservatism and Social Reform. London: Routledge.

Spirling, Arthur and Iain McLean. 2007. “UK OC OK? Interpreting Optimal Classification Scoresfor the United Kingdom House of Commons.” Political Analysis 15(1).

Stokes, S. 2011. “What Killed Vote Buying in Britain?” Unpublished working paper .

Trevelyan, George. 1922. British History in the Nineteenth Century and After (1782–1901). London:Longmans, Green, and Co.

Walker, Brian. 1978. Parliamentary Elections in Ireland, 1801-1922. Dublin: Royal Irish Academy.

29