CANDIDATES OR PARTIES?
Objects of Electoral Choice in Ireland
Michael Marsh
A B S T R A C T
Under many electoral systems, voters can choose between candidates,
and, under some systems, between candidates of the same party – a
situation that makes it possible for candidates to seek a personal vote.
Studies of some countries have shown how personal voting is apparent
in the success of particular types of candidates, notably incumbents, but
there is little systematic study of personal motives among the electors
themselves. The single transferable vote system (STV) used in Ireland
certainly allows electors to choose between candidates as well as parties
and therefore is seen as providing a strong incentive for candidates to
seek personal votes. While aggregate evidence from election results has
pointed to the primary importance of party, survey data have suggested
that close to a majority of voters are primarily candidate-centred. In this
article, an extensive set of instruments contained in the 2002 Irish
election study is used to explore the extent to which voters decide on
candidate-centred factors as opposed to party-centred factors. It is
shown that a substantial minority decide on the basis of candidate
factors, and typical models of Irish electoral behaviour have not accommodated
the heterogeneity that results from this mix of motives.
However, direct questions about motives probably underestimate the
extent of party-centred voting.
KEY WORDS Ⅲ electoral behaviour Ⅲ electoral systems Ⅲ Ireland Ⅲ personal vote
Introduction
Studies of electoral behaviour tend to focus on party choice. When it comes
to high-proﬁle single-candidate elections, such as those for president in the
United States, there is a recognition that the party label is not all that
matters and that personal attributes of the candidates have an importance
independent of party. Yet there is a signiﬁcant and growing literature
PA RT Y P O L I T I C S V O L 1 3 . N o . 4 pp. 500–527
Copyright © 2007 SAGE Publications Los Angeles London New Delhi Singapore
www.sagepublications.com
1354-0688[DOI: 10.1177/1354068807075944]
arguing that candidates themselves should be and are important sources of
votes in many countries and in much less signiﬁcant elections. Candidates
may attract support for who they are, or what they have done, or what they
might do, rather than simply because of the party to which they belong.
There are good institutional reasons for this. Under certain electoral systems,
individual candidates have a strong incentive to differentiate themselves
from others in their party and to develop a personal following. In a widely
cited article, Carey and Shugart (1995) explained how this stimulus would
be higher where the vote was cast for a candidate and not a party and where
that vote had a signiﬁcant effect not just on which parties won seats but on
which candidates did so (see also Katz, 1986; Marsh, 1985b). Many states
use multi-member electoral systems that provide particularly strong incentives,
including Finland, Switzerland and the Irish Republic,1 while many
others, including mixed-member systems such as New Zealand, and singlemember
plurality systems, including Britain, the US and Canada, provide
some encouragement for candidates to seek personal support.
Despite the interest in how electoral systems may lead rational politicians
to develop a personal following, there has been relatively little work
designed to ﬁnd out the extent to which they are able to do this, and much
of that has been by inference – comparing votes won by different politicians
– rather than by direct measurement using voter surveys. For example,
Moser and Scheiner (2005) assess the extent of personal voting in several
mixed-member systems by comparing list and candidate votes for the same
party.2 In single-member district electoral systems there is an extensive
literature looking at how far incumbency seems to confer an advantage and
seeing such effects as indicating a degree of personal voting (e.g. Bean, 1990;
Cain et al., 1987; Gaines, 1998; Krashinsky and Milne, 1986; Wood and
Norton, 1992). Swindle (2002) uses election results to compare levels of
personal voting in Ireland and Japan. He examines the degree of variation
in support for the several candidates of a party within a district and
concludes, perhaps surprisingly, that there is more variation and hence more
personal voting in Ireland. Of course, while such variation may well indicate
that voters do discriminate between candidates from the same party it does
not show whether the vote for the party increases in consequence. It could
well be that personal voting is nested within, and so subsidiary to, party
voting. Taking an even more indirect tack, Shugart et al. (2005) explore the
hypothesis that personal voting is higher in some countries than in others
by identifying symptoms of personal voting: the incidence of candidates
born locally or with local representative experience. Karvonen (2004) also
looks for aggregate symptoms of personal voting, such as higher legislative
turnover and more electoral volatility.
Of Canadian respondents asked to judge the importance of candidates,
leaders, issues and the local candidate in their vote decision, between 20
and 30 percent said the local candidate was the most important factor over
the period 1965–79, which is fewer than chose parties or leaders (Irvine,
M A R S H : C A N D I D AT E S O R PA RT I E S ?
501
1982: 761). The personal qualities of candidates appear to weigh more
heavily in the minds of Finnish voters. Voters were asked whether candidate
was a more important factor in their voting decision than party. Only
a small majority of the respondents said that party was more important
(Raunio, 2004: 5).3
Karp et al. (2002) take a more indirect approach, still using survey data.
They examine the weighting of candidate ratings within a multivariate
model, including ratings of all candidates as well as measures of party
attachments and other variables. They argue that much split-ticket voting
in New Zealand’s mixed-member system is largely the result of personal
voting – a conclusion echoed by Moser and Scheiner (2005). Blais et al.
(forthcoming) have taken this sort of analysis a stage further by simulating
choices in which candidates do not matter and comparing them to actual
ones. They ﬁrst estimate the impact of candidate evaluations on the vote in
Canada and then pose the counterfactual question: how many people would
vote the same way if candidate evaluations were all the same? From their
simulation, using a multivariate model with all candidate evaluations
subsequently set to zero, they conclude that the impact of this would be
small, with only 6 percent of Canadians voting differently. This is in striking
contrast to earlier survey evidence and indicates a much lower level of
personal voting in that country than is suggested by asking people directly
(Irvine, 1982).
Much of this work consists of single country studies. Shugart argues that
comparative work in this area has been limited by the unavailability of
comparative data (2005: 49–50). While there are now extensive data available
on parties and elections there are few on candidates, and hence limited
resources to explore personal voting. Nor is it clear how this can be done in
a way that provides equivalence across countries.4 To suggest that we need
more survey data begs the question of whether, and if so how, surveys can
identify candidate-centred voters. What sort of items might be used to
identify those who cast a personal vote? This article contributes to the wider
literature by providing a detailed case study of possible measures using the
Irish Republic, where elections are fought using a strong preferential voting
system, the single transferable vote (STV) (see below). STV makes candidatecentred
voting compatible with party voting to a degree that is unusual. It
appears to provide a signiﬁcant stimulant to politicians to develop and seek
support on a personal basis but, as Bowler and Farrell point out, ‘while it
may make sense to assume that candidates spend time and effort on “porkbarrel”
and “constituency service” politics, this is no guarantee that this is
the basis for voting behaviour at the level of the electorate’ (1991a, b: 317).
In what follows I show that there is ample evidence that for many voters the
candidate rather than the party is the key to their decision on Election Day.
There has been much discussion on the respective importance of parties
and candidates in Irish elections. Conventional wisdom certainly sees the
personal vote as extremely important. Candidates themselves pander to and
PA RT Y P O L I T I C S 1 3 ( 4 )
502
help to create a demand for personal service and they campaign strongly
for their own personal preference votes, as a number of studies have demonstrated
(see Gallagher and Komito, 2005). There has been less analysis of
the voters themselves and the limited evidence does not tell a coherent story.
Some candidates from a party are more successful than others – in many
cases despite efforts by parties to ensure their support is distributed evenly
(Marsh, 2000; Swindle, 2002). Opinion surveys and exit polls have asked
people about the relative importance of party and candidate in their decision
and the most important factor for up to half of all voters has been the
candidate (King, 2000; Mair, 1987; Sinnott, 1995). This has been underlined
in recent years by the growing success of non-party candidates in
general elections. Even so, there is an obvious conﬂict between the opinion
poll evidence, which suggests that candidate-centred reasons lie behind
many ﬁrst preference votes, and the hard data on election results, which
testify to a considerable stability in party support (Mair and Marsh, 2004).
One problem has been the ambiguity of the opinion poll evidence, not least
because many voters may choose candidates from within a party (Mair,
1987: 92). Until recently, it was not possible to look beyond the evidence
of scattered opinion polls; however, with the ﬁelding of the ﬁrst full election
study in 2002 information is now available to explore the respective weight
of party and candidate much more fully.5
The results of the exploration are signiﬁcant in three ways. Firstly, an
extensive examination using a variety of measures clariﬁes the extent to which
personal voting is prevalent in Ireland, something that is indicated by theory
but not conﬁrmed satisfactorily by the evidence to date. The assessment of
the various types of evidence and measures is also an important step towards
comparative study, since it provides a basis for evaluating different possible
approaches – which themselves may have been developed because of system
variations. In particular, we compare the inferences that can be made from
reported behaviour with assessments by respondents of their own motives.
Secondly, assessing the extent of personal voting is signiﬁcant for our
understanding of the process of electoral democracy. It is common to
assume that the electorate makes parties responsible for government, but it
makes little sense to look for reasons why a particular voter supported a
party if that voter was rather supporting a particular candidate and would
have done so whatever that candidate’s party label. This article thus
examines what (Irish) voters are doing when they vote. In general terms:
are they selecting parties, or are they selecting candidates? If the former is
the case, they could also be voting for a government, or a party leader, but
either way they are behaving in a manner comparable to voters in most
other countries.6 If they are selecting candidates, then our interpretation of
Irish voting behaviour will have to be rather different. This would not be
to conclude that Irish voters are driven by personality. On the contrary, they
could be motivated by concerns about issues and performance in just the
same way that party-centred voters can be, but those concerns would have
M A R S H : C A N D I D AT E S O R PA RT I E S ?
503
to be linked by voters to candidates as individuals, not as representatives of
parties.
Thirdly, identifying the object of electoral choice is important for the ways
in which we explain electoral behaviour. Our explanations normally assume
that voters are thinking about and choosing between parties but, as we have
seen, there is good reason to believe that at least some of them are thinking
about candidates. It is possible that one of the reasons why electoral behaviour
in Ireland is hard to explain using the models developed elsewhere is
that many voters ignore ‘party’. In an often-cited article, Rivers (1990)
warned those exploring electoral behaviour using multivariate models about
the assumption of homogeneity that underlies such models. When heterogeneity
is ignored, the resulting coefﬁcients may well be seriously biased.
While one set of solutions to this heterogeneity has been primarily methodological,
using more appropriate statistical techniques to cope with invalid
assumptions, the main, more substantive, issue raised by Rivers is to identify
the various sources of heterogeneity.
We examine several types of evidence on the relative importance of candidates
and parties in this article using data from the 2002 election study, the
ﬁrst of its kind in the Republic. Firstly, we describe the Irish electoral system
and explain how it promotes candidate-centred voting. Secondly, we examine
how voters actually ﬁll in their ballots. Does the manner in which they do
this suggest that party is the main organizing principle for most voters?
Thirdly, we examine what the voters themselves say about their motivations,
using an open-ended question about their ﬁrst preference vote. Fourthly, we
examine the evidence provided by closed-ended questions about motivations.
Fifthly, using more indirect methods, we examine respondents’ thermometer
ratings of candidates and parties and see how they differ. Who
ranks most highly, the party or the candidate? Each of these methods gives
us a different answer to the question of how extensive candidate-centred
voting is in Ireland. While some differences are small, others are quite
dramatic. In the sixth section of the article, we move beyond simple
categorization and examine the basis for a more nuanced measure of candidate-centredness
based on the various alternatives presented. Finally, we
illustrate the value of this measure by showing that candidate-centredness
is an important source of heterogeneity in Irish voting behaviour.
The main objective here is to assess how far voters focus more on candidates
than on parties, not to explain why they do so. The latter question,
too, is an important one, as the discussion above explains. It is also one that
must be answered if we are to understand political competition, and not just
in Ireland. However, we must ﬁrst ascertain the extent to which people vote
for candidates rather than for parties and compare and evaluate methods of
assessing how important parties and candidates are to each individual voter.
This is the central task of this article. We will also see how conventional
explanations of voting behaviour work much better for voters who appear
to be party-centred than for those who appear more candidate-centred.
PA RT Y P O L I T I C S 1 3 ( 4 )
504
M A R S H : C A N D I D AT E S O R PA RT I E S ?
505
The Irish Electoral System
Application of the single transferable vote (STV) in multi-member constituencies
gives an unusual degree of freedom to the voter to choose between
candidates. The ballot lists the candidates in alphabetical order, indicating
the party of each. To cast a valid vote, the voter must indicate his ﬁrst choice
by placing a ‘1’ next to a candidate’s name. That is sufﬁcient for a valid
vote, but the voter may go on to indicate second, third and later preferences
using the numbers 2, 3 and so on up to the number of candidates on the
ballot. Seats are allocated to candidates who achieve a quota, deﬁned as one
more than the valid vote divided by seats at stake +1. If seats remain unﬁlled
once the ﬁrst preferences have been counted, then there is a further count.
This takes the form either of distributing the ‘surplus’ votes of an elected
candidate or of eliminating the candidate with the fewest votes and distributing
the votes for that candidate over the remaining candidates according
to the next marked preference (Sinnott, 2005).7
Supporters of the STV system point with approval to the fact that voters
may decide on the basis of whatever attributes of the candidates are most
important to them. A voter may be inﬂuenced by party but also by considerations
such as where a candidate lives, that candidate’s gender, or their age
or experience. These are not necessarily exclusive: voters may vote on
locality, for instance, but do so within parties, picking the candidates of a
preferred party according to how close their base is to the voter’s own area.
All this requires information. Voters need to know something about the
candidates, and only party, gender, locality and occupation are sometimes
apparent from the ballot paper.8
The ballot itself does not provide as much help as it might to those who
want to vote on party lines. Certainly it would facilitate, or even encourage,
party voting if it were structured in a series of columns, by party, as it
is in other, similarly preferential, electoral systems, rather than as an alphabetically
ordered long ballot (Darcy and Marsh, 1994). Party names have
been on the ballot since 1965, and these are now complemented by colourful
party logos, but the unwary voter will still have to scan a list of around
a dozen or more candidates carefully if he or she is to organize all their preferences
along party lines.
Making the Ballot
All respondents were provided with something very like the ballot they
would have been faced with on Election Day and were asked to ﬁll it in as
they did at that time.9 Respondents were also offered the option of ﬁlling
in the ballot and placing it in a sealed envelope. Eighty-nine percent of all
respondents and 92 percent of those who claimed to have voted ﬁlled in
the ballot. Using this evidence, we examine how they did so. This involves
scrutinizing not simply ﬁrst preferences but second, third and lower preferences,
and exploring the extent to which people appear to vote for parties
as opposed to candidates. We discuss different ways in which the importance
of the party label might be manifested in the preferences and show how
different deﬁnitions can lead to different conclusions about the role of party.
What sort of pattern would we ﬁnd if party were the dominant criterion
for voters? There has been considerable analysis of the patterns of voting
using the aggregate material at constituency level available from ofﬁcial
results, which indicates a strong degree of voting on party lines, as a high
percentage of votes tend to be transferred between candidates of the same
party (see, for instance, Gallagher, 1978, 1993, 1999, 2003; Marsh, 1981;
Sinnott, 1995). However, the information this gives is limited to those votes
that do transfer. Moreover, the original preference of those voters whose
vote is transferring can soon be lost. Ideally, we would know how each voter
voted and this is what our simulated ballot tells us. In a pioneering analysis
of such data, drawn from European Parliament elections and by-elections,
Bowler and Farrell (1991a, b) discussed how the information from simulated
ballots could shed light on the importance of party (see also King,
2000). The strongest sign that party matters would be that whenever a voter
voted for a candidate he subsequently voted for all the other candidates of
that party in sequence. Party would clearly be the dominant criterion.
Candidate preferences could well matter, but only nested within party preference.
Whether or not this should also be conﬁned to one party is a matter
for debate, but if we apply the criterion all the way down the ballot very
few voters would be classiﬁed as party-centred. A similar, if weaker, sign
would be that all the running mates of the ﬁrst-placed candidate are
supported before any other candidate. Party would be dominant, at least
for the ﬁrst-placed party and for the most inﬂuential preferences in terms
of the outcome.10 Weaker still is a pattern identiﬁed by Laver (2004) in an
analysis of the full record from the three constituencies using electronic
voting in 2002. This is when all of a party’s candidates receive a vote but
not necessarily in sequence. This is signiﬁcant because most voters indicate
a preference for only a few of the candidates standing.11 Ranking two of
the ﬁrst three from a single party, or three of the ﬁrst four, indicates that
party is playing a very strong role.
This general approach infers party-centred voting from a pattern of
preferences that favours a particular party’s candidates. The relationship
between pattern and inference is very deterministic. Voters must adhere to
one of a few candidate orderings in order to be classiﬁed as a party voter.12
Should some allowance be made for random error? In essence, must a party
voter stick strictly to the party list or can there be some deviation? If so,
how great a deviation? Considering a party voter as one who votes, a
complete ticket may be a more realistic basis for deﬁnition than requiring
that he do so in a strict sequence. A second problem is that where a party
runs a single candidate it is not possible to see any difference between a
PA RT Y P O L I T I C S 1 3 ( 4 )
506
M A R S H : C A N D I D AT E S O R PA RT I E S ?
507
party vote and a personal vote. Someone who picks the sole Labour candidate
as No. 1, the sole Green as No. 2 and sole Progressive Democrat as
No. 3 may be voting on party grounds but may also be choosing on some
other basis. We cannot tell simply on the basis of the simulated ballot. We
can analyse voting patterns where a party ﬁelds more than a single candidate
and try to generalize from that situation to others. This is not wholly
unproblematic since most multi-candidate situations involve either Fianna
Fáil or Fine Gael and their voters may be more loyal, more party-centred,
than those of other parties. However, some contrasts between Fianna Fáil,
Fine Gael and Labour are apparent and will be considered when generalizations
are made to the wider electorate.
Using several deﬁnitions, the patterns of voting are examined in Table 1.
Analysis in this section is conﬁned to parties running more than one candidate
in the respondent’s constituency. It shows the proportions of voters
who cast a complete vote, i.e. support all the candidates of their ﬁrstpreference
party. Once they cast their ﬁrst-preference vote for a party, 60
percent vote for all remaining candidates of that party. The ﬁgure is a little
lower for Labour than for Fianna Fáil and Fine Gael, but the differences
are quite small. Most such complete ballots are also sequential. Forty-eight
percent of those giving their ﬁrst preference to Fianna Fáil cast a complete
and sequential ballot – akin to the classic ‘straight-ticket’. The ﬁgure is
lower for the other parties, but more than two out of every ﬁve supporters
of those parties cast a complete and sequential ballot. Forty-four percent of
those casting their ﬁrst preference for a party running more than one candidate
cast a vote for all that party’s candidates before expressing any other
preferences. Moreover, many of the departures from a strict sequence are
small, involving the interpolation of a single candidate. Overall, most ﬁrstpreference
votes for parties translate into votes for the whole party slate, and
the majority of the latter are cast in sequence. While this still allows for a
considerable degree of candidate-centredness within the party slate, it does
imply that party is the most important element for a large number of voters.
Table 1. Patterns of voting in multi-candidate situations
Fianna Fine
Fáil Gael Labour All
Voting a complete party list – as a % of ﬁrst 62 60 50 60
preferences
Voting a complete party list in sequence – 48 40 36 45
as a % of ﬁrst preferences
First two votes for candidates from same 61 46 41 55
party – as a % of ﬁrst preferences
N 964 410 84 1472
Notes: Includes only instances where a party ﬁelded more than one candidate. Self-reported
voters only. Those voting for independents and others are excluded.
PA RT Y P O L I T I C S 1 3 ( 4 )
508
Bowler and Farrell (1991a, b) suggest another way to look at the inﬂuence
of party on the way people ﬁll in the ballot. This involves an examination
of the extent to which voters cast a vote for two successive candidates
of the same party. Each preference set can be seen as a number of pairs – 1
and 2, 2 and 3, 3 and 4 and so on – and each can be seen as a pair from
either the same party or from different parties. A strong partisan structure
would show a high number of party-consistent pairs; a weak structure
would show a low number or none at all. This offers a potentially more
nuanced measure of party-centredness using a summation of the number of
party pairs, but voters give different numbers of preferences and again there
is the problem that many parties run no more than one candidate in any
constituency, and that different constituencies have different numbers of
party-consistent options. Limiting analysis to just the ﬁrst pair, the third row
in Table 1, shows that in 2002 the majority of voters who could do so voted
for candidates from the same party with their ﬁrst- and second-preference
vote. Fifty-four percent voted for two candidates of the same party.13
Whatever deﬁnition is employed, we have demonstrated that while some
voters act as if their choice is strongly party-centred, some do not; seemingly,
an attribute of the candidate other than party is a critical one for
many. How large the proportion of voters is that we might describe as candidate-centred
rather than party-centred is a question of deﬁnition. On some
counts, well over half of all voters may be termed party-centred.
Reasons for First-Preference Vote
A second way of estimating the relative importance of candidate and party
in vote choice is to use a direct question: to ask respondents themselves to
explain their vote choice. Respondents were asked a number of questions
about their choice. We asked: ‘Thinking about the candidate you gave your
ﬁrst-preference vote to, what was the main reason you voted for that particular
candidate rather than any other candidate?’ This was followed by ‘And
what other reasons did you have for giving your ﬁrst-preference vote to that
candidate?’ In grouping the answers to that question, interest is particularly
in the breakdown between those that made the party of that candidate
central and those that emphasized some other aspect of that candidate.
Most answers fall within one of four categories: personal characteristics
of the candidate, the area the candidate comes from, the party of the candidate
and the candidate’s policies. The ﬁrst set is essentially personal: the
voter knows the candidate, the candidate is ‘good’, the ‘best candidate’ is
‘honest’ or ‘sincere’. This is not to say that performance does not matter:
many see the candidate as a ‘good worker’ or a ‘hard worker’, or as someone
who has been ‘helpful’ to the voter. The second set of answers highlights
local representation: the candidate is from the area, or has been good for
the area, and has a ‘good record’ in the area or is a ‘good worker’ for the
M A R S H : C A N D I D AT E S O R PA RT I E S ?
509
area. The third set is essentially partisanship, giving the party of the candidate
as the key reason. Finally, there are ‘policy’ justiﬁcations, citing the views
or opinions of the candidate. Other reasons include a view that the candidate
represented particular interests (farmers, workers, business or the elderly),
tactical or strategic voting, and vague reference to family factors that are
not clearly either personal or party. Table 2 gives the distribution of these
motives across the sample, and shows, ﬁrstly, the main reason given and,
secondly, all reasons, including subsidiary ones. When asked, people appear
to see candidates in terms of who they are and what they have done rather
than their party or policy. Half of all respondents who gave any reason
provided an essentially personal justiﬁcation and only one in ﬁve spontaneously
mentioned party. However, it is obvious that in some instances a
respondent might feel partisanship would be an inappropriate answer.
Anyone voting for one of the two, three or four Fianna Fáil candidates in
a constituency, for instance, might feel the need to explain why they chose
that Fianna Fáil candidate rather than another. The same could be true of
most Fine Gael voters and many of those voting Labour.
In the second part of Table 2 we show the responses of those who voted
for a candidate who had no running mate from the same party. The main
differences between this group and the whole are a decrease in those giving
personal responses, as we would expect, and in those stressing area representation.
This is compensated by a striking increase in those giving a policy
response and a small increase in those giving a party response. The implication
of this may be that area and personal factors are more important for
selecting within parties than between them, while policy is more important
for selecting between parties. However, it is also possible that these factors
vary across parties. By conﬁning analysis to single-candidate situations we
reduce the impact of Fianna Fáil and Fine Gael voters on the distribution
and increase the weight of those voting Green or Sinn Féin. We also examined
the responses by the party of the candidate receiving the ﬁrst preference and
Table 2. Main and subsidiary reasons offered for selecting ﬁrst choice candidate
All parties Single candidate parties only
Main reason All reasons Main reason All reasons
% % % %
Personal attributes 39 51 36 41
Party 20 27 27 33
Policy 8 11 18 24
Area represented 30 35 20 26
Other 11 16 13 17
N 1880 1880 655 655
Note: The sum of each column may exceed 100 due to multiple responses. Up to three main
reasons and three subsidiary reasons coded for each respondent.
Self-reported voters only. Those voting for independents and others are excluded.
by the numbers of candidates from that party. This conﬁrms not only that,
indeed, responses vary considerably within parties according to how many
candidates are running, but also that they vary between parties, even when
allowing for numbers of candidates. The example of Fine Gael, which has
sufﬁcient instances of one, two or more candidates, shows that party is less
likely to be given as a reason where there are multiple Fine Gael candidates.
This analysis suggests that most voters appear to be attracted by the
personal characteristics and attributes of the candidates themselves rather
than by their party. However, it is arguable that this evidence understates
the importance of party. As we have seen, the number of candidates put up
by a party has an impact on the way the question is answered. There are few
instances of single-candidate situations for Fine Gael and none for Fianna
Fáil. It may also be that ‘party’ is a response which may be unacceptable to
many who feel parties have a low reputation in general and that they will
appear more conscientious if they can give apparently more informed reasons
for their support. Finally, the stress in the question on the choice of candidate
may also have encouraged respondents to provide non-partisan answers.
However, even if the party-centred voter is not as rare as Table 2 suggests,
these answers do highlight the signiﬁcance of factors other than party and
to that extent reinforce the ﬁndings in the earlier section which indicated
that many voters did not appear to vote as if motivated primarily by party.
They also reveal the cultural norms of voting and representation that devalue
party and emphasize personal and local service.
Closed-ended Questions about First Preference Vote
In addition to the open-ended questions, we asked people directly to tell us
whether party or candidate was most important for their decision on ﬁrst
preference.14 Only 39 percent responded by selecting party, the majority
saying it was the candidate that was most important (Table 3).15 Party is
most important for Greens, Sinn Féin and, to a lesser extent, Fianna Fáil.
We later posed the same question in a different way, asking respondents if
they would still have voted for the same candidate had that candidate stood
for a different party (Table 4).16 There is considerable consistency at the
individual level across the two questions, with only 17 percent giving apparently
inconsistent answers: claiming to be candidate-centred but saying they
would not follow the candidate into a different party, or claiming to be
party-centred but willing to follow the candidate into a different party. The
results are also similar in the aggregate, with 38 percent saying they would
not follow a candidate who changed party, as against 39 percent saying that
party (as opposed to candidate) was the major factor in their choice.
Those who said ‘it depends’ are drawn almost equally from those who
previously gave party and those who gave candidate as the main reason for
their ﬁrst preference vote. Neither response pattern is logically inconsistent.
PA RT Y P O L I T I C S 1 3 ( 4 )
510
M A R S H : C A N D I D AT E S O R PA RT I E S ?
511
Party-motivated voters who would nonetheless follow a candidate could
view two (or more) parties with almost equal approval. Candidate-motivated
voters might still see party as a factor sufﬁcient to limit their choice. One
respondent explained his ﬁrst preference vote by saying his favoured candidate
was ‘anti-Fianna Fáil’. This underlines the point that both candidate
and party factors should be assumed to play a role for any voter, but that
while for some the weighting of the two may be equal, for others the weighting
is very unequal. As it becomes more unequal it is reasonable to classify
voters as primarily ‘party-centred’ or as primarily ‘candidate-centred’, but
this does not mean other considerations are entirely absent. It again appears
that party is weighted most strongly among those who support the Greens,
Sinn Féin and Fianna Fáil candidates and weakest among those supporting
candidates from Fine Gael, Labour and the Progressive Democrats.
Combining the two measures, we ﬁnd that 38 percent are clearly candidatecentred,
26 percent party-centred and 36 percent not unambiguously of
either type.17 Green and Sinn Féin voters are most party-centred; Fine Gael
and Progressive Democrat voters are most candidate-centred (see Table 5).
Table 3. Self-reported most important factor in deciding ﬁrst preference, party or
candidate?
Candidate Party Total N
% %
Fianna Fáil 56.0 44.0 100.0 947
Fine Gael 64.6 35.4 100.0 443
Green 41.5 58.5 100.0 84
Labour 68.0 32.0 100.0 200
PD 72.0 28.0 100.0 70
Sinn Féin 44.7 55.3 100.0 97
Total 58.7 41.3 100.0 1841
Self-reported voters only. Those voting for independents and others are excluded.
Table 4. Would respondent vote for same candidate if candidate stood for
different party?
Yes Depends No
Party % % % Total N
Fianna Fáil 42.7 14.7 42.6 100.0 958
Fine Gael 58.5 10.7 30.8 100.0 451
Green 15.8 26.0 58.2 100.0 87
Labour 44.8 22.5 32.7 100.0 201
PD 48.8 24.2 26.9 100.0 70
Sinn Féin 37.3 19.2 43.5 100.0 99
Total 46.0 15.8 38.2 100.0 1866
Self-reported voters only. Those voting for independents and others are excluded.
PA RT Y P O L I T I C S 1 3 ( 4 )
512
Candidate and Party Ratings
Asking people to voice their reasons for making any choice is problematic,
because people may in reality make decisions without thinking through the
sorts of criteria they are being asked to consider. They may also be unaware
of the way in which certain predispositions impact on their evaluations of
the choices offered. While these direct questions are useful they should not
be seen as deﬁnitive. Ranking is also less than ideal, as we do not know
whether a primary reason clearly outweighs a secondary one, or whether
the margin is a small one. Questions that ask respondents to rate several
things on a scale can be more nuanced and so more useful. Much electoral
research is based on asking people to rate stimuli – leaders, issues, performances
and so on – on a number of scales and the most important factors in
vote choice are then inferred from the pattern of correlations. This may be
a simple enough exercise where there are only two or three parties; it is
much more time-consuming where there may be up to 17, as is the case with
candidates. However, we asked our respondents to rate each of the parties
and each of the candidates from those parties on a thermometer scale.18
We feared that independent and minor party candidates would be particularly
difﬁcult for respondents to evaluate and so excluded such candidates
from this part of the survey. If voters are to decide on the basis of candidates
rather than parties, it is necessary for them to know something about the
candidates other than their party label. Seventy-ﬁve percent of all voters
report having met the candidate to whom they gave their ﬁrst preference and,
more importantly, the rating data demonstrate that most voters seem able to
differentiate between many candidates. They do not know something only
about their ﬁrst choice. In fact, respondents were generally willing to evaluate
Table 5. Candidate or party index from direct questions
Candidate-centred Mixed Party-centred
% % % Total N
Fianna Fáil 35.0 35.0 30.0 100.0 964
Fine Gael 49.8 26.8 23.5 100.0 451
Green 15.0 41.6 43.4 100.0 87
Labour 40.5 36.7 22.8 100.0 204
PD 46.7 35.0 18.2 100.0 70
Sinn Féin 25.4 34.4 40.2 100.0 99
Total 38.7 33.0 28.3 100.0 1875
Candidate-centred voters are those whose candidate is the primary factor in their ﬁrst preference
vote and that if their candidate had run for some other party they would still have voted
for him. Party-centred voters are those who say party is the primary factor in their ﬁrst
preference vote and that if their candidate had not run for that party they would not have
voted for him. The rest, including non-respondents to one or other question, are classiﬁed
as ambivalent.
Self-reported voters only. Those voting for independents and others are excluded.
M A R S H : C A N D I D AT E S O R PA RT I E S ?
513
the majority of candidates. More than half rated all of them and the average
respondent rated more than 70 percent of candidates. This compares with
95 percent who rated all parties.
Of course, it remains to be seen how far party and candidate are differentiated
and it is the relative ratings of parties and candidates that we are interested
in most. It might be expected that party-centred voters would rate party
above, or at least equal with, that party’s candidates and that candidatecentred
voters would rate candidates more highly. A clear majority of respondents
do differentiate candidate and party, with 62 percent giving a rating
to the party of the candidate who gets their ﬁrst preference vote that is different
from that they give to the candidate. Table 6 gives the average rating of
the ﬁrst-preference candidate and the average rating of the party of that
candidate, again broken down by party.
Table 6 demonstrates that the average voter rates his ﬁrst-preference
candidate more highly than he rates his ﬁrst-preference party. The difference
is small enough, but it is signiﬁcant at the 0.01 level. This is generally true
across all parties, with the exception of the Greens and Fianna Fáil, where
the average party rating is slightly higher than the average candidate rating.
Within these parties the difference is very small: only in the case of Fianna
Fáil is it signiﬁcant at even the 0.05 level. However, the pattern is similar to
those we have observed above, with party evaluations higher than candidate
evaluations for Green, Fianna Fáil and Sinn Féin voters and lower than candidate
evaluations among voters supporting Fine Gael, Progressive Democrat
and Labour candidates.19
We can also use this indirect measure to explore lower preferences. This
is given in Table 7 for the ﬁrst three preferences. Looking ﬁrst at all voters
in column 1 we see that candidates obtaining a ﬁrst preference are rated
more highly than their parties, but that the situation is reversed for those
obtaining lower preferences. (This remains true even when we conﬁne the
analysis to those who expressed at least three preferences.) However, this
pattern varies substantially between those whose vote appears to be more
Table 6. Candidate and party ratings for respondents’ ﬁrst preferences
Mean rating of Mean rating of
ﬁrst preference ﬁrst preference Mean individual
candidate candidate’s party difference
Fianna Fáil 80.0 81.4 –1.4
Fine Gael 79.4 68.8 +10.2
Green 72.3 74.8 –1.9
Lab 76.6 69.4 +6.8
PD 77.5 70.2 +7.8
Sinn Féin 82.6 77.6 +3.9
All 79.3 76.2 +2.9
Self-reported voters only. Those voting for independents and others are excluded.
PA RT Y P O L I T I C S 1 3 ( 4 )
514
or less party-centred. In column 2, where analysis is limited to those casting
a sequential ballot, those who vote for all candidates of their ﬁrst-preference
party in sequence, party rates higher than candidate for all preferences,
although there is a big gap between 1 and 2. In contrast, when we look only
at those who do not cast a sequential ballot, mean candidate ratings exceed
those for party and are similar for both ﬁrst and second preferences. Limiting
analysis only to those who cast a party-inconsistent set of ﬁrst and second
preferences, we see the same is true: candidates outweigh party but the
relative ratings are similar for the ﬁrst two candidates.
This demonstrates that party is a stronger determinant of lower preferences
than it is of ﬁrst preferences, although for some voters there is little
difference between the ﬁrst and the second preference in the primary importance
of candidate. This general primacy of party over candidate can be
understood in terms of information, as Richardson (1988) has argued in the
case of Japan. In what is a small political context, most voters do have good
information about perhaps one or two candidates, but after that they know
more about parties than candidates, and judge accordingly.
Comparing the Measures of Candidate-/Party-centred
Voting
We have examined several measures of candidate-centred voting. Before we
go further, we brieﬂy review these measures and compare the estimates given
by each measure of the extent to which Irish voters are primarily candidateor
party-centred. Comparison is conﬁned here to the group of voters analysed
in Table 1: those who did not vote for an independent and who could vote
for at least two candidates of their (apparently) preferred party. First ballot
behaviour: what proportion voted for all the candidates of the same party
sequentially and what proportion voted at least for all the candidates of that
party, even if not in strict sequence. The ﬁrst group – comprising 44 percent
– we see as essentially party voters, the second – comprising 16 percent –
as having well-mixed motives and the residual 42 percent as candidatecentred.
Our pair of closed-ended questions when combined yields a lower
Table 7. Candidate–party differences for lower preferences
All voters All voters
casting not casting All voters casting
sequential sequential party-inconsistent
All voters ballot ballot ﬁrst pair
First preference +3.3 –2.5 +6.5 +10.1
Second preference –2.3 –13.1 +5.2 +8.8
Third preference –4.1 –10.8 +0.4 –2.8
Self-reported voters only. Those voting for independents and others are excluded.
M A R S H : C A N D I D AT E S O R PA RT I E S ?
515
ﬁgure for party-centred voters and a higher one for those with mixed orientations.
Comparing candidate and party rankings again gives a smaller
ﬁgure for party-centred voters, with only 27 percent favouring party as
opposed to 35 percent favouring the candidate. However, only the openended
question seems to elicit a very different response, with just 20 percent
giving party as a main reason and a further 7 percent giving it as a contributory
reason. Seventy-three percent do not mention party at all. The problems
with the open-ended question that may account for the low number of
apparently party-centred voters have already been identiﬁed. Discounting
the open-ended measurements, it seems reasonable to say that around twoﬁfths
of voters appear to be essentially candidate-centred, while the rest
incline more to party. Interestingly, however, the behavioural measure gives
the highest estimate for more narrowly party-centred voters, while the other
measures suggest no more than one-quarter of voters are ﬁrmly partycentred.
It seems that inferences from reported behaviour of voters will give
a higher ﬁgure for party-centredness than inferences from reported motives,
although, judging by the results of Blais et al. (forthcoming) discussed above,
it is possible that even the former may underestimate the real importance
of party factors. I revisit this point below.
There is not space here to explore in any detail who the more candidatecentred
voters are.20 The key point is that there is evidence based on a
variety of methods that suggests importance of party in the voting decision
varies considerably within the Irish electorate. In the last section of this
article I explain the implications of this for our explanations of Irish voting
behaviour.
Explaining Electoral Choice
Here, I examine the performance of what we argue is a typical multivariate
model of party choice across different types of voters, differentiated by the
extent to which they are candidate- rather than party-centred. Before we
Table 8. Candidate- versus party-centred ﬁrst preference voting: a comparison of
distributions obtained using different measures
Candidate-centred Mixed Party-centred
Ballot behaviour1 42 16 44
Closed questions2 38 36 26
Open question: ‘party’ response3 75 6 19
Ratings 35 37 27
Notes: Includes only instances where a party ﬁelded more than one candidate. Self-reported
voters only Those voting for independents and others are excluded.
1 Mixed includes those who cast a complete but not a sequential ballot.
2 See Table 6.
3 Mixed includes those for whom ‘party’ was a stated reason but not the main reason.
PA RT Y P O L I T I C S 1 3 ( 4 )
516
discuss the model we must explain how such a differentiation is made. We
could simply pick one of the several measures discussed above. However,
on the assumption that these measures all tap the same phenomenon, it
makes more sense to combine them into a single measure. This can be done
using factor analysis, which also gives us some indication of how far each
of the measures reﬂects the same phenomenon. If they do, the factor analysis
will indicate that a single dimension could underlie all ‘responses’. Table 9
displays the results of two principal factor analyses. The ﬁrst includes almost
all of the items identiﬁed in Tables 1–6,21 the second those which loaded
reasonably well (in practice at 0.39 and above) on the ﬁrst dimension in the
ﬁrst analysis. To enable the analysis to include all voters for parties (those
for independents and ‘others’ are excluded), the two behavioural measures
were coded at 0 when the voter’s ﬁrst choice had no running mates. In
addition, the ratings measure was reversed to make party dominance a
positive score: in Table 6 it is a negative score.
The eigenvalue for the ﬁrst factor is just over 2.0, indicating clear evidence
of a common factor, if not a strong one. With the exception of most coded
responses to the open-ended question, all items load at 0.39 or above, and
the ‘party’ response to the open-ended question also loads at 0.40. The policy
response is almost completely unconnected to the primary dimension. A
second analysis, excluding the three weakest items, produces a generally
better solution, with the eigenvalue almost unchanged, although the openended
‘party’ response now loads at only 0.31. This gives us some conﬁdence
Table 9. Principal factor analyses of candidate-/party-centred voter measures
Factor loadings
on ﬁrst factor:
Factor loadings unrotated
on ﬁrst factor: Weakest measures
unrotated excluded
Open-ended: Policy .05 –
Open-ended: Party .40 .31
Open-ended: Personal –.21 –
Open-ended: Area –.14 –
Voted consistent ﬁrst pair .70 .74
Voted sequential party list .71 .75
Direct: party versus candidate .60 .57
Direct: hypothetical .58 .56
Indirect: party versus candidate .39 .37
Eigenvalue 2.03 1.97
Notes: N = 1389. Each variable codes 1, 0, –1; those voters who gave ﬁrst preferences to a party
running only one candidate and who thus could not be classiﬁed as party- or candidatecentred
on the basis of the ballot are treated as an intermediate category (0), as are ties on
the indirect measure and those who answered ‘it depends’ on the hypothetical measure.
Those voting for independents and others are excluded.
that these measures do reﬂect the same phenomenon. The alpha index
measure of reliability for these ﬁve items is a reasonable 0.60 (0.62 without
the open-ended measure).22
We can now use this derived measure of party-/candidate-centredness –
the factor scores – to examine the need for a different explanation for the
electoral behaviour of party- and candidate-centred voters. Explanations of
electoral behaviour generally emphasize party, something candidate-centred
voters see as relatively unimportant. The analysis here is intended to be no
more than illustrative. We use a conventional model explaining electoral
behaviour.23 Our expectation is that the model will work poorly in explaining
the behaviour of more candidate-centred voters and much better with
respect to other voters. The model contains the following variables:
• Demographics: (non-)membership of union, membership of the Gaelic
Athletic Association, gender, urban–rural location (all 1, 0), age in years
and education (1–6).
• Ideological position: self-placement on the national issue, left–right and
abortion (11-point scales).24
• Party attachment (1, 0, with ‘leaners’ at 0).25
• Evaluations of the performance of the economy, 3-point scale scored –1
to +1; attribution of credit or blame to government (1, 0) and interaction.26
• Evaluations of party leaders (11-point scales).
The estimates for this model for the 2002 election are given in Table 10.
In line with most other analyses of Irish electoral behaviour it is clear that
social cleavage measures are weak predictors of voting (Laver, 1986; Marsh
and Sinnott, 1993, 1999; Whyte, 1974). Ideological issues too are weak (see
Laver et al., 1988; cf. Bowler and Farrell, 1990) and so is the performance
of the economy – perhaps because it was doing so well. The most signiﬁcant
factors are partisanship and comparative assessments of leaders (Carty,
1981; Harrison and Marsh, 1994). Overall, the McFadden adjusted R2 for
the model is 0.31, hardly a high ﬁgure given the large number of variables
included. It is accepted here that this model might be improved; different
measures of government performance, alternative scales of political values
and perhaps more sensitive measures of social status might all lead to more
signiﬁcant coefﬁcients and an improvement overall in the adequacy of the
model. However, this model is typical enough of those used to explain choice
in many countries. Following Rivers (1990), we should be very cautious
about any of these estimates, since ignoring existing heterogeneity might
lead us to overestimate or underestimate actual effects. And, of course, it is
argued here that this model is severely affected by heterogeneity because
many voters are not strongly inﬂuenced by party-related cues.
Our contention is that the weakness of the model is in part a function of
the strong candidate-centred norm among Irish voters. One crude way to
test this is to run the model separately for relatively candidate-centred and
for relatively party-centred voters. For this distinction, we have divided the
M A R S H : C A N D I D AT E S O R PA RT I E S ?
517
PA RT Y P O L I T I C S 1 3 ( 4 )
518
factor scores from the second analysis in Table 9 at the median. The respective
McFadden adjusted R2 ﬁgures for the two models are 0.11 (N = 640) for
relatively candidate-centred voters and 0.51 (N = 740) for relatively partycentred
voters. This is a considerable difference but, of course, the basis
for estimation becomes smaller in the two groups and the variances differ.
Moreover, this distinction between party and candidate is very crude.
A better way to test our contention is to compare the performance of the
overall model for several different subgroups of voters by examining the
prediction errors. We can do this most easily by comparing the predictions
given by the model against actual party choice and seeing whether predictions
are more accurate for more party-centred respondents. Such a comparison
Table 10. Multinomial logit estimation of vote choice model
PDs Fine Gael Greens Labour Sinn Féin
Leader evaluations
Ahern (Fianna Fáil) –.055*** –.041*** –.042*** –.051*** –.040***
Harney (PD) .042*** .002 –.002 –.006 –.015+
Noonan (Fine Gael) –.003 .034*** .004 –.004 –.009
Sargent (Green) –.016+ –.013** .078** –.015* –.013
Quinn (Lab) .015 .008 .002 .047*** .007
Adams (Sinn Féin) .005 –.006 –.008 .005 .065***
Party attachment
Fianna Fáil –1.19* –1.61*** –1.98+ –4.35*** –3.62**
Progressive Democrat 1.53+ .286 –.440
Fine Gael 1.04 3.12*** –.081
Green .842 1.98+ 2.38+ –3.75+ 3.17+
Labour .538 .667 . – .3.90*** .341
Sinn Féin 1.30 .781 3.10+ 5.92***
Ideological position
Abandon United Ireland –.026 –.014 .038 –.055 –.098
Spend and tax more –.056 .021 –.134+ .042 .071
Pro-abortion –.044 –.031 –.065 .010 –.019
Government Performance
(a) The economy better .224 .371 –.774 .736 .006
(b) Govt responsible for
(b) economy –.702 –.031 –.035 .613 –.444
(a) * (b) .630 –.611+ –.847 –1.15* –.227
Demographics
Rural versus urban –.928** .389+ –.314 –.542* –.721*
GAA member –.219 –.060 –.584 –.607+ –.621
Union member –.641+ .235 –.506 –.140 .194
Age in years .011 –.001 .012 .001 .045***
Woman –.379 .098 .314 .114 .135
More education .125 .109 .303* .212* –.051
Constant –22.47 2.88 –27.11 –1.81 88.89**
Note: McFadden’s adj R2 = 0.32; N = 1401; log-likelihood = –1186. Voters for independent
candidates excluded. The reference category is Fianna Fail. ***Signiﬁcant at 0.001;
**signiﬁcant at 0.01; *signiﬁcant at 0.05; +signiﬁcant at 0.10. Cells are blank when there
is no variance.
M A R S H : C A N D I D AT E S O R PA RT I E S ?
519
is displayed in Figure 1 for voters of all parties (the solid line) as well as for
those of Fianna Fáil (dashed line) and Fine Gael (dotted line) alone. A perfect
model would predict actual ﬁrst-preference votes with complete certainty
(i.e. with a probability of 1.00). If our contention that this model will be more
effective in predicting the votes of more party-centred voters is correct, then
the prediction of actual ﬁrst preference will come closer to 1.00 as the voter
is more party-centred. We have again used the factor scores from the second
analysis given in Table 9 to measure degrees of party-centredness, and
rounded them to give us a 5-point scale from –2 (most candidate-centred)
to +2 (most party-centred). As Figure 1 shows, the average prediction of ﬁrstpreference
vote for the most party-centred voters is almost 0.8; for the least
party-centred voters it is 0.2. Predicted probabilities across the six possible
outcomes in Table 10 (including Fianna Fáil, the reference category) will
always sum to 1.00, so 0.8 is a satisfactory performance and 0.2 is clearly
not at all good. The fact that the results for the voters of Fianna Fáil and
Fine Gael, respectively, are similar in shape indicates that this pattern is not
simply an artefact, because a Fianna Fáil vote is on average most likely and
Fianna Fáil supporters as a whole seem party-centred, as we saw above. For
Fianna Fáil the slope runs from a low of 0.40 to a high of 0.80, while for
Fine Gael it runs from 0.12 to 0.86 and for the full sample from 0.14 to
0.80, an increase in accuracy of 500 percent.27
Figure 1. Predicted probability of actual ﬁrst preference vote. (Voter type
measured by factor scores from analysis in Table 9: independent voters and those
supporting others are excluded)
–2 –1 0 1 2
Meanpredictionofactualvote
00.20.60.40.8
Candidate–party scale
All parties FF voters
FG voters
PA RT Y P O L I T I C S 1 3 ( 4 )
520
Discussion and Conclusions
In this article, I set out to examine the extent of personal voting in Ireland.
While there is consensus that many Irish electors are strongly candidatecentred,
the systematic evidence to support the claim and assess the relative
importance of candidate and party has been absent. The measures that
have been used in the past have seemed to indicate different answers and
contain sufﬁcient uncertainty or ambiguity to caution us against drawing
ﬁrm conclusions.
There are many ways to explore how important party is to voters. We have
chosen four here: open-ended questions, a simulated ballot, direct closedended
questions and indirect scale measures. All coincide in indicating a
very signiﬁcant degree of candidate-centred voting; the more direct measures
suggest most and the more indirect ones suggest least. All concur in identifying
some parties’ voters as more party-centred (Greens, Sinn Féin, Fianna
Fáil) and others (Labour, Progressive Democrats, Fine Gael) as more candidate-centred,
at least in 2002. Several measures do suggest that around 40
percent of voters for these parties are signiﬁcantly candidate-centred.28 In
general, the measures based on behaviour indicate lower levels of candidatecentred
voting than those based on reported behaviour, while open-ended
questions seem to indicate by far the highest levels. It seems likely that
although many voters may vote a party ticket they will rationalize this to
themselves in terms of candidate qualities. The open-ended measure correlates
least well with the other measures, but including all types of measure in
a factor analysis still results in an acceptable scale. The cutoff points of this
would be arbitrary, so this does not tell us how many voters are either partyor
candidate-centred, but it does provide a measure of differences in degree.
Behavioural measures are not applicable across many electoral contexts.
However, the results here indicate that the sort of closed-ended questions
used can produce results that are reasonably equivalent to those obtained
using reported behaviour. This should be helpful in comparative work.
The apparent paradox of candidate-centred voting combined with partisan
stability at aggregate level is hardly resolved by this analysis. Of course, the
paradox may be no more than apparent. The party system is becoming much
more unstable – as we might expect in the absence of strong party anchors
in the electorate. Survey evidence points to less party identiﬁcation (Marsh,
2006) and aggregate evidence is that levels of partisan solidarity in voting
patterns are lower than ever (Gallagher, 2003: 105–6), while independent
candidates have done unusually well in recent elections. The next Irish
election study will provide important evidence on the association between
candidate-centred voting and partisan stability as it will complete a panel
started at the 2002 election. The data will show how far voters who were
more party-centred in 2002 actually do remain more faithful to their parties,
and show how far candidate-centred voters whose candidate is no longer
standing are nonetheless attracted to another candidate of the same party.
It is also possible that the importance of candidate remains overestimated
by all the measures used here. Candidate evaluations may be added to the
model estimated in Table 10 and each respondent’s predicted vote choice
calculated, both with candidate evaluations included and again with all
candidate evaluations set to zero. The two sets of predictions can then be
compared, as Blais et al. (forthcoming) have done for Canada. We did this
and concluded that only 20 percent of voters would have made a different
party choice (with Fine Gael, as might be expected, the biggest loser).29 This
is a simple simulation that pays no attention to the standard errors that
surround any such prediction. Nor does it allow for the fact that parties
may be liked because of a candidate they nominate. Even so, while the result
suggests that the personal vote is more important in Ireland than in Canada,
it does at least warn us that candidate-centred voting may be less extensive
than is suggested in Table 8.
It has also been argued here that differences in the degree to which Irish
voters are party-centred indicates a heterogeneity in the electorate that has
not been modelled by any analysis to date. This applies whether 20 percent
or 40 percent of voters are candidate-centred. While it has been recognized
that different Irish voters probably do use different criteria when deciding
how to vote, the precise implications of this for multivariate models have
not been pursued. We have seen here that the more candidate-centred the
voter, the less easily is his or her vote predicted by a conventional model of
electoral behaviour, one that emphasizes party. This has obvious implications
for our understanding of how people vote and what the vote means.
While lip service is frequently paid to the idea that there are 42 separate
contests (one in each constituency), too often the signiﬁcance of this is
ignored when commentators interpret the decision of the electorate. Commentators
and academics may draw policy lessons from the vote where they are
not justiﬁed, since many voters may have chosen their candidate for reasons
quite unconnected with that candidate’s party. The lesson of this for the study
of Irish electoral behaviour is that the research agenda for Irish electoral
behaviour must include the development of candidate-centred assessments,
paralleling those of party, which can be integrated into models of the vote.
The same surely applies in other countries where individual candidates matter.
Notes
I thank John Garry, Fiachra Kennedy and Richard Sinnott for comments on an
earlier draft of this article. Versions of the paper have been presented at departmental
seminars in Trinity College Dublin and at the universities of Aberdeen and
Trondheim, as well as at APSA in 2003 and the Political Studies Association of
Ireland meeting in 2005. I am grateful to all participants for their suggestions. I am
also grateful to the Irish Council for Research in the Humanities and Social Sciences
for a senior research scholarship 2002–03 which gave me time to write the initial
draft of the article.
M A R S H : C A N D I D AT E S O R PA RT I E S ?
521
PA RT Y P O L I T I C S 1 3 ( 4 )
522
1 A recent review identiﬁed 10 states with such strong preferential voting (Chile,
Cyprus, Estonia, Finland, Greece, Liechtenstein, Luxembourg, Poland, San
Marino and Switzerland) plus a further four with candidate votes that are not
automatically pooled at party level (Ireland, Malta, Mauritius and Vanuatu)
(Karvonen, 2004: 208). Shugart also includes Brazil and Peru, but sees Estonia’s
list system as more strongly determined by party ordering. In addition, recent
changes have increased the importance of the preferential vote in Austria and
Belgium (Shugart, 2005: 41–3).
2 Rejecting earlier arguments that this demonstrated strategic voting, Moser and
Scheiner argue that it indicates a personal vote for that candidate. This is because
the candidate–party difference for the ﬁrst-placed candidate is positively correlated
with the closeness of the race and not negatively correlated, as would be
implied by strategic voting. They show this is the case in several systems although
not in Germany, a result they attribute to the high level of party institutionalization
in that country.
3 See also the 2004 Finnish election study results for question 72: http://www.
fsd.uta.ﬁ/english/data/catalogue/FSD1260/cbf1260e.pdf.
4 Module 2 of the Comparative Studies of Electoral Systems (CSES) project did
include questions on candidate recognition and candidate contact and, where preferential
voting was possible, whether the respondent cast a candidate preference:
see http://www.umich.edu/~cses
5 This study was funded under the Programme for Research in Third Level Institutions
of the Higher Education Authority of the National Development Plan
and directed by Michael Marsh and Richard Sinnott. The data are available from
the Irish Social Science Data Archive: http://www.ucd.ie/issda/.
6 Dalton and Wattenberg (2000) argue that this situation itself is changing, but
their interpretation of party-centred voting is narrower than that employed here.
7 As the most highly ranked candidates in terms of ﬁrst preferences win the great
majority of seats, success seems largely unaffected by lower preferences. However,
this holds only because of the extent to which the distributions of lower preferences
are broadly similar to those of ﬁrst preferences. Also, at most elections there
are a number of candidates who owe their election to winning more second and
third preferences than ﬁrst preferences: see Gallagher (1978, 1979).
8 In this context, a voter might well use ballot position as a cue, and there is certainly
some evidence for this (Marsh, 1987; Robson and Walsh, 1974).
9 The simulated ballots lacked the candidate photographs and party logos of the
real thing, though they did feature party names. They thus resembled closely the
pre-2002 style of ballot.
10 Bowler and Farrell (1991a) called this pattern an ‘unravelling’ one, in the sense
that preferences for a single party’s candidates come ﬁrst, followed by a mixture
of candidate and party preferences for the less signiﬁcant votes.
11 Laver (2004) analysed actual preference data from three constituencies that voted
electronically in the 2002 Irish general election. His results appear broadly similar
to those from our simulated ballot with respect to the number of preferences
expressed and the partisan patterning. The median number of preferences was
3 in Laver’s analysis and here. The mean number of preferences was between
4.4 and 5 across the three constituencies Laver examined, as opposed to 4 in the
election study data. The number we have per constituency is less than 100, but
for the three constituencies analysed by Laver mean preferences are lower, being
M A R S H : C A N D I D AT E S O R PA RT I E S ?
523
between 3.1 and 3.7. Our respondents were certainly much less likely to complete
a full ballot. However, comparing the proportion of Fianna Fáil and Fine Gael
ﬁrst-preference voters casting a complete and sequential ballot for Fianna Fáil
(Table 1) with Laver’s ﬁgures in his tables 2–3, results are similar. According to
Laver’s results, between 46 and 53 percent of Fianna Fáil voters gave complete
and sequential votes, and between 39 and 43 percent of Fine Gael voters. This
goes some way to validating the use of simulated ballots.
12 There are parallels here in the literature on attitude scaling, particularly on
cumulative scales (see, e.g., McIver and Carmines, 1994).
13 As Bowler and Farrell (1991a, b) discovered, consistent pairs are much more
common at the top of the ballot, with almost three-quarters of all such pairs
being those between ﬁrst and second preference and between second and third
preference. Examining all voters who cast at least two preferences, the proportions
casting consistent pairs each time are 37 percent, 25 percent, 17 percent,
14 percent and 9 percent, respectively, for the ﬁrst ﬁve pairings.
14 The measures in this section come closest to questions found in opinion polls
over many years. These ask respondents which of several options best describes
the reasons for their choice. Most of the reasons are party- and policy-related,
but one is ‘picking the best candidate to represent the needs of the constituency’.
It is the latter that has been chosen by between 40 and 50 percent of respondents
over the past 20 years (see Mair, 1987; Marsh and Sinnott, 1999; Sinnott, 1995).
15 The question is: ‘Which would you say was more important in deciding how you
cast your ﬁrst-preference vote in the general election in May of this year – the
party or the candidate him/herself?’
16 The question is: ‘If this candidate had been running for any of the other parties
would you still have given a ﬁrst-preference vote to him/her?’
17 Party-centred voters are those who say party is the primary factor in their ﬁrstpreference
vote and that if their candidate had not run for that party they would
not have voted for him. Candidate-centred voters are those for whom candidate
is the primary factor in their ﬁrst-preference vote and who say that if their
candidate had run for some other party they would still have voted for him. The
rest are classiﬁed as ambivalent.
18 ‘I’d like to ask you how you feel about some Irish politicians, using what we call
the “feeling thermometer”. The feeling thermometer works like this: If you have
a favourable feeling (a warm feeling) towards a POLITICIAN you should place
him/her somewhere between 50 and 100 degrees; If you have an unfavourable
feeling (a cold feeling) towards a POLITICIAN you should place him/her
somewhere between 0 and 50 degrees; if you have a don’t feel particularly warm
or cold (have no feeling towards the politician at all), then you should place
him/her somewhere at 50 degrees; where would you place these Irish politicians?
And where would you place each of the following PARTIES; and where would
you place these CANDIDATES who ran in your constituency in the general
election in May?’
19 Different voters may use the thermometer scale in different ways despite the
instructions they are given (see Brady, 1985). This can be countered by calculating
party and candidate scores for each respondent that are centred on each
respondent’s mean party and candidate score. However, this standardized measure
may be affected by the fact that some parties ﬁeld more candidates than others,
and if all candidates from a favoured party are highly rated the difference
PA RT Y P O L I T I C S 1 3 ( 4 )
524
between the top candidate and the candidate mean will then be smaller. For this
reason we have simply used the unstandardized measures here.
20 In fact, candidate-centred voters are more likely to be young, female, rural and
less knowledgable about politics, although none of these bivariate associations
are strong ones.
21 ‘Complete’ voting was excluded, as this is similar to the sequential vote measure
and the aim is rather to look at similarities between rather different measures.
22 A problem with the open-ended measure is that it places so few people in the
party category, for the reasons already discussed.
23 For multivariate models of Irish voting, see e.g. Garry et al. (2003), Mair (1987),
Marsh (1985a), Marsh and Sinnott (1999), McAllister and O’Connell (1984).
24 These are bipolar scales. The pairs are: insist on a united Ireland now – abandon
the aim of a united Ireland altogether; government should cut taxes a lot and
spend much less on health and social services – government should increase taxes
a lot and spend more on health and social services; there should be a total ban
on abortion in Ireland – abortion should be freely available in Ireland to any
woman who wants to have one.
25 Party attachment is measured using the CSES Wave 2 question. ‘Leaners’ are
people who think of themselves as ‘close to a political party’ but, when asked how
close, say they are ‘not very close’, or who say only they are closer to one party
than the others.
26 ‘Thinking back over the past ﬁve years – the lifetime of the 1997 to 2002 Fianna
Fáil/Progressive Democrat government – would you say the ECONOMY in
Ireland over that period of time got a lot better, a little better, stayed the same,
got a little worse or got a lot worse?’ Five response options ranged from ‘Got a
lot better’ to ‘Got a lot worse’, plus a sixth option, ‘don’t know’. This was
followed by: ‘Do you think this was MAINLY due to the policies of that government
or NOT MAINLY DUE to the policies of that government?’
27 The correlation between the raw factor scores measuring candidate versus party
from Table 8 and predicted probabilities of voting Fianna Fáil among those who
actually did so is 0.89; it is 0.92 for Fine Gael, 0.90 for Labour, 0.83 for the
Greens, 0.91 for the PD, 0.91 for Sinn Féin and 0.90 for all parties together. All
correlations are highly signiﬁcant. Overall, for a shift of 1 standard deviation
towards a more party-centred vote, the probability of a correct prediction
increases by 0.09.
28 Of course this excludes those who voted for non-party candidates, 9.5 percent
in 2002.
29 We included in the model a set of evaluations for each respondent indicating the
best evaluation of a candidate from each party, i.e. best Fianna Fáil candidate,
best Fine Gael candidate and so on. All evaluations were centred on zero (rather
than 50) and 0 was made the middle point in all other scales. The ‘prediction’
is made by taking the most likely outcome for each voter from a complete model
and then comparing this with a set of ‘predictions’ made from a new dataset in
which all candidate evaluations are set to zero. Our analysis differs from that of
Blais et al. (forthcoming) in that they use multinomial probit while we used
multinomial logit. Our model was also more extensive.
M A R S H : C A N D I D AT E S O R PA RT I E S ?
525
References
Bean, C. (1990) ‘The Personal Vote in Australian Federal Elections’, Political Studies
38: 253–68.
Bowler, Shaun and David M. Farrell (1990) ‘Irish Voter Rationality: The 1987 Irish
General Election Revisited’, Economic and Social Review 21: 251–68.
Bowler, Shaun and David Farrell (1991a) ‘Voter Behaviour under STV-PR: Solving
the Puzzle of the Irish Party System’, Political Behaviour 13: 303–20.
Bowler, Shaun and David Farrell (1991b) ‘Party Loyalties in Complex Settings: STV
and Party Identiﬁcation’, Political Studies 39: 350–62.
Blais, André, Elizabeth Gildengil, Agnieszka Dobrzynska, Neal Nevitte and Richard
Nadeau (forthcoming) ‘Does the Local Candidate Matter?’ Canadian Journal of
Political Science.
Brady, Henry E. (1985) ‘The Perils of Survey Research: Interpersonally Incomparable
Responses’, Political Methodology 11: 269–91.
Cain, B., J. Ferejohn and M. Fiorina (1987) The Personal Vote: Constituency Service
and Electoral Independence. Cambridge: Harvard University Press.
Carey John M. and Matthew S. Shugart (1995) ‘Incentives to Cultivate a Personal
Vote: A Rank Ordering of Electoral Formulas’, Electoral Studies 14: 417–39.
Carty, R. K. (1981) Party and Parish Pump: Electoral Politics in Ireland. Waterloo,
ON: Wilfrid Laurier University Press.
Dalton, Russell and Martin P. Wattenberg (2000) Parties without Partisans: Political
Change in Advanced Industrial Democracies. Oxford: Oxford University Press.
Darcy, R. and Michael Marsh (1994) ‘Decision Heuristics: Ticket Splitting and the
Irish Voter’, Electoral Studies 13: 38–49.
Gaines, B. J. (1998) ‘The Impersonal Vote? Constituency Service and Incumbency
Advantage in British Elections, 1950–92’, Legislative Studies Quarterly 23:
167–95.
Gallagher, Michael (1978) ‘Party Solidarity, Exclusivity and Inter-Party Relationship
in Ireland 1922–1977: The Evidence of Transfers’, Economic and Social Review
10: 1–22.
Gallagher, Michael (1979) ‘The Impact of Lower Preference Votes on Irish Parliamentary
Elections, 1922–1977’, Economic and Social Review 11: 19–32.
Gallagher, Michael (1993) ‘The Election of the 27th Dáil’, in Michael Gallagher and
Michael Laver (eds) How Ireland Voted 1992, pp. 57–78. Dublin: PSAI Press.
Gallagher, Michael (1999) ‘The Results Analysed’, in Michael Marsh and Paul
Mitchell (eds) How Ireland Voted 1997, pp. 121–50. Boulder, CO: Westview
Press.
Gallagher, Michael (2003) ‘Stability and Turmoil: Analysis of the Results’, in
Michael Gallagher, Michael Marsh and Paul Mitchell (eds) How Ireland Voted
2002, pp. 88–118. London: Palgrave.
Gallagher, Michael and Lee Komito (2005) ‘The Constituency Role of TDs’, in John
Coakley and Michael Gallagher (eds) Politics in the Republic of Ireland, 4th edn,
pp. 242–71. London: Routledge.
Garry, John, Fiachra Kennedy, Michael Marsh and Richard Sinnott (2003) ‘What
Decided the Election?’ in Michael Gallagher, Michael Marsh and Paul M. Mitchell
(eds) How Ireland Voted 2002, pp. 119–42. London: Palgrave.
Harrison, Michael and Michael Marsh (1994) ‘What Can He Do for Us?’, Electoral
Studies 13: 289–312.
PA RT Y P O L I T I C S 1 3 ( 4 )
526
Irvine, William P. (1982) ‘Does the Candidate Make a Difference? The MacroPolitics
and Micro-Politics of Getting Elected’, Canadian Journal of Political
Science 15: 755–82.
Karp, Jeffrey, Jack Vowles, Susan A. Banducci and Todd Donovan (2002) ‘Strategic
Voting, Party Activity, and Candidate Effects’, Electoral Studies 21: 1–22.
Karvonen, Lauri (2004) ‘Preferential Voting: Incidence and Effects’, International
Political Science Review 25: 203–26.
Katz, Richard S. (1986) ‘Intraparty Preference Voting’, in Bernard Grofman and
Arendt Lijphart (eds) Electoral Laws and Their Consequences, pp. 85–103. New
York: Agathon Press.
King, Simon (2000) Parties, Issues and Personalities: The Structural Determinants
of Irish Voting Behaviour from 1885 to 2000. Thesis. University of Oxford.
Krashinsky, M. and W. J. Milne (1986) ‘The Effect of Incumbency on the 1984
Federal and 1985 Ontario Elections’, Canadian Journal of Political Science 19:
337–43.
Laver, Michael (1986) ‘Ireland: Politics with Some Social Bases: An Interpretation
Based on Survey Data’, Economic and Social Review 17: 193–213.
Laver, Michael (2004) ‘Analysing Structures of Party Preference in Electronic Voting
Data’, Party Politics 10: 521–41.
Laver, Michael, Michael Marsh and Richard Sinnott (1988) ‘Patterns of Party
Support’, in Michael Laver, Peter Mair and Richard Sinnott (eds) How Ireland
Voted 1987, pp. 99–140. Dublin: Poolbeg/PSAI Press.
Mair, Peter (1987) The Changing Irish Party System: Organisation, Ideology and
Electoral Competition. London: Frances Pinter.
Mair, Peter and Michael Marsh (2004) ‘Political Parties in Electoral Markets in
Postwar Ireland’, in Peter Mair, Wolfgang Muller and Fritz Plasser (eds) Political
Parties and Electoral Change: Party Responses to Electoral Markets, pp. 234–63.
London: Sage.
Marsh, Michael (1981) ‘Localism, Candidate Selection and Electoral Preference in
Ireland’, Economic and Social Review 12: 167–86.
Marsh, Michael (1985a) ‘Ireland’, in Ivor Crewe and David Denver (eds) Electoral
Change in Western Democracies: Patterns and Sources of Electoral Volatility,
pp. 173–20. London: Croom Helm.
Marsh, Michael (1985b) ‘The Voters Decide? Preferential Voting in European List
Systems’, European Journal of Political Research 13: 365–78.
Marsh, Michael (1987) ‘Electoral Evaluations of Candidates in Irish General
Elections, 1948–1982’, Irish Political Studies 2: 65–76.
Marsh, Michael (2000) ‘Candidate Centred but Party Wrapped: Campaigning in
Ireland under STV’, in Shaun Bowler and Bernard Grofman (eds) Elections in
Australia, Ireland and Malta under the Single Transferable Vote, pp. 114–30. Ann
Arbor, MI: Michigan University Press.
Marsh, Michael (2006) ‘Party Identiﬁcation in Ireland: An Insecure Anchor for a
Floating Party System’, Electoral Studies 25: 489–508.
Marsh, Michael and Richard Sinnott (1993) ‘The Voters: Stability and Change’, in
Michael Gallagher and Michael Laver (eds) How Ireland Voted 1992, pp. 93–114.
Dublin, Folens and Limerick: PSAI Press.
Marsh, Michael and Richard Sinnott (1999) ‘The Behaviour of the Irish Voter’, in
Michael Marsh and Paul Mitchell (eds) How Ireland Voted 1997, pp. 151–80.
Boulder, CO: Westview Press.
M A R S H : C A N D I D AT E S O R PA RT I E S ?
527
McAllister, Ian and Declan O’Connell (1984) ‘The Political Sociology of Party
Support in Ireland: A Reassessment’, Comparative Politics 16: 191–204.
McIver, John P. and Edward G. Carmines (1994) ‘Unidimensional Scaling’, in Michael
S. Lewis-Beck (ed.) Basic Measurement, pp. 139–228. International Handbooks of
Quantitative Applications in the Social Sciences, vol. 4. London: Sage.
Moser, Robert and Ethan Scheiner (2005) ‘Strategic Ticket Splitting and the Personal
Vote in Mixed-member Electoral Systems’, Legislative Studies Quarterly 30:
259–76.
Raunio, Tapio (2004) ‘2004 European Parliament Election Brieﬁng No. 16: The
European Parliament Election in Finland, June 13, 2004’, (European Parties
Elections and Referendums Network: Sussex University). Available at: http://
www.sussex.ac.uk/sei/documents/epernep2004ﬁnland.pdf
Richardson, Bradley M. (1988) ‘Constituency Candidates versus Parties in Japanese
Voting Behaviour’, American Political Science Review 82: 695–718.
Rivers, Douglas (1988) ‘Heterogeneity in Models of Electoral Choice’, American
Journal of Political Science 32: 737–57.
Robson, Christopher and Brendan Walsh (1974) ‘The Importance of Positional Bias
in the Irish General Election of 1973’, Political Studies 12: 191–203.
Shugart, Matthew S. (2005) ‘Comparative Electoral Systems Research: The Maturation
of a Field and New Challenges Ahead’, in Michael Gallagher and Paul
Mitchell (eds) The Politics of Electoral Systems, pp. 25–56. Oxford: Oxford
University Press.
Shugart, Matthew S., Melody Ellis Valdini and Kati Suominen (2005) ‘Looking for
Locals: Voter Information Demands and Personal Vote-Earning Attributes of
Legislators under Proportional Representation’, American Journal of Political
Science 49: 437–49.
Sinnott, Richard (1995) Irish Voters Decide: Voting Behaviour in Elections and
Referendums Since 1918. Manchester: Manchester University Press.
Sinnott, Richard (2005) ‘The Rules of the Electoral Game’, in John Coakley and
Michael Gallagher (eds) Politics in the Republic of Ireland, 4th edn, pp. 105–34.
London: Routledge.
Swindle, Stephen M. (2002) ‘The Supply and Demand of the Personal Vote’, Party
Politics 8: 279–300.
Whyte, John (1974) ‘Ireland: Politics Without Social Bases’, in R. Rose (ed.) Electoral
Behaviour: A Comparative Handbook, pp. 619–52. New York: Free Press.
Wood, D. M. and P. Norton (1992) ‘Do Candidates Matter? Constituency-Speciﬁc
Vote: Changes for Incumbent MPs, 1983–1987’, Political Studies 42: 227–38.
MICHAEL MARSH is an Associate Professor of Political Science in Trinity College
Dublin and Head of the School of Social Sciences and Philosophy. He has published
widely in the ﬁeld of electoral behaviour, was a principal investigator for the 2002
Irish Election Study and is principal investigator for the follow-up study in 2007
and co-author of The Irish Voter (Manchester University Press, 2007).
ADDRESS: Department of Political Science, Trinity College Dublin, Republic of
Ireland. [email: mmarsh@tcd.ie]
Paper submitted 13 July 2005; accepted for publication 18 November 2005.