Content uploaded by Tom Louwerse
Author content
All content in this area was uploaded by Tom Louwerse on Jun 30, 2015
Content may be subject to copyright.
The impact of opinion polls on party mandate
fullment: evidence from Ireland, the Netherlands and
the United Kingdom
Tom Louwerse
∗
Trinity College Dublin
Abstract
What is the eect of opinion polls on mandate fullment? This paper studies
political parties' policy congruence between their pre-electoral manifesto and parlia-
mentary speech. The expectation is that congruence depends on parties' perform-
ance in recent opinion polls, conditional on a party being in government or in oppos-
ition. The paper provides empirical evidence from Ireland, the Netherlands and the
United Kingdom (1997-2015). Based on Wordscores estimates of parties' pre- and
post-electoral position, some support is found for the hypothesis that government
parties show increased congruence the more their opinion poll support changes, while
opposition parties show (marginally) decreased congruence.
Many studies have shown that the extent to which parties full their pre-electoral pledges
or programmes is imperfect, but better than generally thought. Variation in pledge full-
ment is related to the electoral system used, ministerial control over portfolios, inclusion
in coalition agreements as well as consensus on specic policies (Klingemann, Hoerbert
and Budge, 1994; Louwerse, 2012; Thomson et al., 2012). All of these explanations are
related to the extent to which parties can implement their programme in a four or ve
year term of government. The reality of policy formation is, however, dynamic: parties
in government and in parliament are subject to a variety of pressures (for example from
civil servants, public opinion, lobbyists) and might take their electoral mandate more
seriously during certain parts of the electoral cycle than others. To fully understand the
role that the party mandate plays in policy making, we need to explore these dynamics
of policy-making.
This paper is part of a project that seeks to combine the party mandate literature with
the literature on policy responsiveness (Wlezien and Soroka, 2012; Stimson, Mackuen and
∗
Paper presented at the European Political Science Association (EPSA) Conference, Austria, Vienna,
25-27 June 2015. Parts of this paper are derived from an earlier draft, presented at the ECPR General
Conference, Glasgow, 2014. This paper is part of the project 'Between Mandate and Responsiveness:
How Electoral Uncertainty Aects Political Representation' which is supported by the Arts and Social
Sciences Benefactions Fund 2013-14, Trinity College Dublin, Ireland.
1
Erikson, 1995). Whereas the former looks at the degree to which pre-electoral commit-
ments are honoured, the latter is concerned with the degree to which governments (or
parliamentary parties) respond to changes in public opinion. While both might be argued
to be relevant from the perspective of political representation, they can oer conicting
demands to political parties, potentially raising discontent with both mechanisms of policy
representation. This paper examines one aspect of the relationship between mandate and
responsiveness by examining the impact of opinion polls on the party mandate.
The paper looks at the congruence of parties pre-electoral commitments in terms of
the policy positions in their manifestos and parliamentary speeches, measured on the left-
right scale. Thus, it looks at parties'
representative
mandate rather than their
government
mandate. I compare three countries that are dierent in terms of institutional design
and the level of mandate fullment: Ireland, the Netherlands and the United Kingdom
(Mansergh and Thomson, 2007; Lijphart, 2012).
The party mandate
The party mandate model of political representation requires parties to present their
(policy) plans before elections and act upon them afterwards. Three approaches have
been taken to study the extent to which parties `full' their `mandate': the pledge ap-
proach, saliency approach and spatial approach (Klingemann, Hoerbert and Budge,
1994; Louwerse, 2011
b
; Royed, 1996). The pledge approach measures whether parties
full their specic election pledges in government. The saliency approach looks at the
correspondence between policy priorities in manifestos and spending priorities after elec-
tions. The spatial approach looks at parties' policy positions on issue dimensions before
and after elections. The last approach bears most resemblance to how congruence between
citizens and representatives is usually studied: by comparing policy positions on issue di-
mensions (Powell, 2000; McDonald and Budge, 2005; Golder and Stramski, 2010). The
advantage is that one can evaluate mandate fullment even when the political agenda
changes.
Most studies of the party mandate look at whether parties are able to translate pre-
electoral commitments into government action: their mandate for government. Alternat-
ively, one might think of the party mandate as a
representative
mandate: a guideline to
how party representatives should act in the legislative arena (Louwerse, 2011
a
; Dalton,
Farrell and McAllister, 2011, 7). Both perspectives are valuable; the added value of
studying parties' representative mandate is that one can evaluate both how government
and opposition parties behave. While opposition parties might be unable to translate
their wishes into government action, we should still expect them to behave in line with
their pre-electoral programme. Therefore, mandate fullment is dened in terms of the
congruence between parties' pre-electoral and parliamentary issue positions.
2
Table 1: Overview of Hypotheses
Type of eect
Similar for government
and opposition Dierent for
government and
opposition
Type of
poll change
Up vs. down Losers Drift
Hypothesis Conditional Losers
Drift Hypothesis
Absolute
change Poll Change
Hypothesis Conditional Poll
Change Hypothesis
Opinion polls and the party mandate
The party mandate model brings a great democratic promise: that voters are able to
inuence policy-making by choosing between dierent policy programmes. In this view,
parliamentary representation and governance should be about translating pre-electoral
stances into actual policy. While previous research shows that mandate fullment, albeit
imperfect, is generally higher than citizens seem to think (Mansergh and Thomson, 2007;
Louwerse, 2012; Naurin, 2002), parliamentary parties and government ministers are sub-
ject to a range of pressures that might impact their ability and willingness to implement
their pre-electoral policy positions. Some of these pressures are related to the institutional
setting, such as the type of executive (presidentialism versus parliamentarism) and the
type of government (i.e. single party majority, coalition, single party minority), and are
therefore constant across a parliamentary term. For example, in terms of the enactment
of specic pledges, coalition systems are at a disadvantage (Thomson, 2001). Other pres-
sures relate to factors that can change over the course of a parliament, such as pressure
from within the bureaucracy, interest groups (lobbying) as well as the overall popularity
of the government.
This paper focuses on the impact of parties' standing in opinion polls on mandate
fullment. It builds on the idea that most governments will be interested in their re-
election (Strøm, 1990, 1994). That is, we should expect parties not only to pursue certain
policies, or to enjoy the spoils of oce, but also to consider their chances at the next
general elections (Soroka and Wlezien, 2010, 137). After all, their ability to obtain policy
or oce in the future does depend on their electoral fortunes. While motivations of party
(leaders) are complex, the assumption that party leaders will be willing to change their
policy positions to some extent in order to gain electoral support seems not too far-fetched
(Hobolt and Klemmensen, 2008). Somer-Topcu (2009) nds support for the hypothesis
that parties change their manifesto positions more if they lost during the last elections.
Similarly, we might expect that party leaders are willing to change their stance, during a
parliamentary term, based on opinion poll results.
The literature on policy
responsiveness
nds some support for the idea that when
government parties' polling margins decline, they will become more responsive to pub-
lic opinion (Soroka and Wlezien, 2010; Hobolt and Klemmensen, 2008). This eect was
particularly strong in the United States, although Hobolt and Klemmensen also found
3
support for this mechanism in Britain (rhetorical responsiveness) and Denmark (eective
responsiveness). If parties become more responsive to the public wishes during a par-
liamentary term, however, they might feel the need to deviate from their pre-electoral
stance. One argument is that a party's election mandate and public opinion will be
(partially) at odds. At least some of a party's policy positions will not be in line with
public opinion - even if that party won a majority in the election, as the Ostrogorski
paradox shows (Thomassen, 1994). If this is the case a parliamentary party has to choose
between fullling its electoral mandate and responding to public opinion. In that case,
we would expect that increases in a party's poll numbers, and with that its electoral
prospects, would increase its mandate fullment. But if a party grows more unpopular, it
will reposition itself to match public opinion more closely, at the expense of its mandate:
Losers Drift Hypothesis
Opinion poll losses decrease congruence between manifesto
and parliamentary policy positions
Alternatively, one might theorize that any change in polls, whether a win or a loss, will
lead to a deviation from the party mandate. For parties that are losing electoral support,
the above mechanism of nding a more suitable electoral position may come into play. But
for winning parties one might like-wise argue that they would not exactly stick to their
manifesto position. Increasing support in the polls may provide parties with the feeling of
certainty about their election prospects and therefore the idea that they can aord a bit of
leeway. After all, parties are not only interested in winning the next elections, but also in
policy and oce (Strøm, 1990). Just as parties that are gaining in the polls might be less
responsive to public opinion, they might be more willing to deviate from their manifesto.
After all, while election manifestos can be seen as an `authoritative' statement of party
policy preferences, they are not necessarily a `pure' reection of the leadership's genuine
policy preferences. Manifestos are set in a strategic environment taking into account
voters' and other parties' preferences (Budge et al., 2001, 77; Thomson, 1999, 3-5; Marks
et al., 2007, 26-27). Moreover, there might be internal disagreement with regard to some
of the policies pursued in the manifesto. Leaders might even believe that certain policies
are the `responsible' (Mair, 2009) thing to pursue, which is easier if the party is doing
well in opinion polls, one might argue. Taken together, we might argue that any change
in opinion polls (both comparing to the election result and quarter-on-quarter) will result
in decreased congruence:
Poll Change Hypothesis
Absolute changes in opinion poll scores decrease congruence
between manifesto and parliamentary policy positions
The hypotheses formulated above can be argued to be conditional upon parties position in
government or opposition. Previous research has demonstrated that this distinction has
a substantive impact on manifesto-parliamentary policy position congruence (Louwerse,
2012; Fivaz, Louwerse and Schwarz, 2014). The (majoritarian) logic of the `responsible
party model' dictates that parties that failed to win the election are `out' and therefore
need to nd a new policy stance in order to appeal to voters at the subsequent elections
4
(APSR, 1950; Ranney, 1954). While this argument does not apply in all settings, espe-
cially when there are multiple opposition parties (Louwerse, 2011
b
), one could make the
argument that at least opposition parties have more of an incentive to adjust their policy
positions, especially if they are losing in the polls. Similarly, government parties regularly
see the necessity to abandon or dilute their pledges because of coalition politics, interest
group pressure or opposition from within government departments. If that leads to a loss
in popular support in the polls, government parties might remedy this by moving back
towards their manifesto position:
Conditional Losers Drift Hypothesis
Opinion poll gains decrease congruence between
manifesto and parliamentary policy positions for opposition parties, but increase
congruence between manifesto and parliamentary policy positions for government
parties
Again, we might consider an alternative explanation, which states that the eect of polls
is not so much in the winning or losing but rather in the absolute change (see Table 1).
Parties that see changes in their electoral fortunes might respond to this by changing
their policy position more than parties that are relatively stable in terms of electoral
support in polls. This absolute `Poll Change' eect can be argued to be conditional upon
being in government or in opposition. When opposition parties are faced with changes
in their electoral position, this might motivate them to pursue a dierent set of policies.
For government parties, changes in their electoral fortunes might, however, signal either a
need to `return to base', certainly for coalition government parties that have compromised
to get into government, or if things go well in the polls they might be able to use this as
leverage to stick to their manifesto policies in the face of opposition from societal groups
or government departments:
Conditional Poll Change Hypothesis
Absolute changes in opinion poll scores de-
crease congruence between manifesto and parliamentary policy positions for oppos-
ition parties, but increase congruence between manifesto and parliamentary policy
positions for government parties
Case selection, data and methods
The analysis focuses on the Ireland, the Netherlands and the United Kingdom, three
countries have been found to dier substantially with regard to pledge fullment, thus
oering diversity in observations (Mansergh and Thomson, 2007). Ideally the above
hypotheses would be tested on a broad range of country cases, but limitations in the
availability of parliamentary data in an accessible fashion necessitate a non-random case
selection strategy. The Netherlands and Ireland both have electoral systems that result in
proportional representation (and usually in coalition government), but the style of politics
is consensual in the former and more adversarial in the latter. Moreover, politics in Ireland
does to a lesser degree follow ideological dierences, with both main parties being on the
5
centre right. The United Kingdom is a majoritarian democracy, often featuring single
party governments. Thus, the hypotheses are tested in diverse settings (Seawright and
Gerring, 2008).
Policy congruence is operationalized in terms of the relationship between a party's
manifesto and its parliamentary position on the left-right scale. The advantage of using
parliamentary speech is that it provides a relatively detailed account of the policy stances
taken by parties in parliament. This makes it well suited for the purpose of tracking
parties' policy stances over the course of a parliament. I treat each quarter, dened as each
period of 91 days since the election, as a time point. This leaves about 16 to 20 quarters
(four to ve years) per full parliamentary term, although the Dutch governments often
failed to complete their term in oce. We focus on the period between 1997 (Netherlands:
1998) and 2015, which includes 4 parliamentary terms for Ireland and the United Kingdom
and 6 for the Netherlands
1
.
Left-right policy positions are estimated through a Wordscores analysis of party mani-
festos and parliamentary speeches given by party members (Laver, Benoit and Garry,
2003). I use party manifestos as reference texts to estimate parliamentary positions.
While one should be careful in comparing dierent types of political texts, in this case
manifestos and parliamentary speech, Warwick (2014) suggests that this actually does
provide reasonable estimates of party positions. The main problem the researcher is
presented with when applying Wordscores is that estimates of virgin texts `shrink' to-
wards the mean because of common words that are plentiful in all texts analysed (Laver,
Benoit and Garry, 2003, 316). Two transformations have been proposed: the Laver,
Benoit and Garry (LBG) transformation, which assumes that the standard deviation of
virgin and reference texts is equal, and the Martin and Vanberg (MV) transformation
which assumes that the `shrinkage' is equal for reference texts and virgin texts (Martin
and Vanberg, 2007; Warwick, 2014). The latter assumption is problematic when using
dierent types of texts: when using manifestos as reference texts, shrinkage was found
to be smaller in manifestos than in parliamentary debates, most likely because manifesto
state policy positions more concisely. For example, when estimating positions in the 2012
Dutch manifestos and subsequent parliamentary term, using 2006 and 2010 manifestos
as reference texts, untransformed scores showed a range of 0.71 for the manifestos, but
only 0.13 for the positions based on parliamentary debates. While we might expect some
moderation of policy positions in parliament, it seems unreasonable to assume that party
manifesto positions are ve times more polarized than their parliamentary positions. The
LBG transformation assumes that reference and virgin texts show the same degree of
variance. Warwick (2014) rejects this approach for his purposes, which is to estimate
government policy positions in budget speeches based on manifesto reference texts, as it
might sweep any real moderation in budget speeches under the carpet. For the purposes
of the current paper, the LBG assumption seems, however, less problematic, as I am
comparing policy positions of all parties before and after elections. As we will see, the
1
The current terms in Ireland (2011-now) and the Netherlands (2012-now) are included up to the start
of 2015. The 2010-2015 House of Commons in Britain is fully included.
6
analysis provides reasonable estimates of party positions in the parliamentary debates.
Source data for texts were obtained from the following sources. Irish and British
manifestos were downloaded from the Polidocs archive, save for the most recent British
manifestos, which were gathered from party websites (Benoit, Bräuninger and Debus,
2009). Dutch manifestos were obtained from published collections with recent manifes-
tos downloaded from the party websites (Pellikaan, Voerman and van Holsteyn, 2003;
Pellikaan et al., 2006). Manifestos that were not available in plain text format, were
converted; page numbers, headers and footers, tables of contents and tables with numer-
ical data were removed. For the Irish manifestos, any Irish-language text was deleted.
The parliamentary debates were obtained from the Ocial Reports of the Houses of
the
Oireachtas
(Oce of the Houses of the Oireachtas, 2014, 2015), the Parliamentary
Parser Project for the UK (Parliamentary Parser Project, 2015) and
Ociële Bekend-
makingen
for the Netherlands (Ociële Bekendmakingen, 2012). Additional data on MP
aliations was obtained from the database of the Parliamentary Documentation Centre
(Parlementair Documentatie Centrum, 2010). All documents were stemmed and cleaned,
i.e. changed to lower case, numbers removed and common stopwords as well as party
indicators removed.
Chapel Hill Expert Survey estimates of parties' positions on the left-right scale are
used to obtain the word scores (Bakker et al., 2012). To arrive at a more robust set of word
parameters, I include the set of manifestos from the previous and subsequent elections
as well, e.g. for the 2005 British elections, the 2001 and 2010 manifestos were included
as well. The next step is to re-estimate the manifesto positions of the year of interest
using these word parameters, which corrects for any impact the inclusion of previous and
subsequent manifestos might have
2
. Finally, the parliamentary positions were estimated,
where the set of speeches by MPs for one party during one quarter was included as a
single text
3
.
Polling data for the United Kingdom is based on the weekly ICM polls for
The Guard-
ian
(Guardian Data, 2014), while for the Netherlands polls from
Ipsos
and its predecessors
were used (partly obtained directly from the company and more recent polls gathered from
their website by the author). For Ireland a set of opinion polls from various pollsters was
used, collected by the author from pollsters' websites as well as the
Irish Opinion Poll
Archive
(Marsh, 2015). Multiple polls in a quarter where averaged. I use four indicators
of poll support. The rst two are the change in party support since the last elections
and its absolute value. The second pair consists of the change in party support since the
previous quarter and the absolute value of that.
Parties for which manifesto, parliamentary voting data and polling data are available
are included in the analysis. This means that for the United Kingdom Labour, Con-
servatives and Liberal Democrats are included; for Ireland Fine Gael, Labour, Fianna
2
If this step is omitted the parliamentary documents are estimated using a dierent set of word scores
than the manifestos of interest.
3
Short sets of speeches in a quarter, for example quarters that overlapped heavily with the summer
recess, were excluded from the analysis as these estimates are less robust. The criterion used is that the
length of the speeches should be at least 50% of the mean speech length per quarter for that party.
7
Table 2: Descriptive statistics
Variable
n ¯x s
Min Max
Party parliamentary position 1072 5.1 2.1 -0.8 11.8
Party parliamentary position
t-1
1022 5.2 2.2 -3.7 12.7
Party manifesto position 1072 5.1 1.9 1.4 10.8
Poll change since election
t-1
939 -0.3 4.6 -24.4 16.2
Absolute poll change since election
t-1
939 3.1 3.4 0.0 24.4
Poll change since last quarter
t-1
935 -0.1 1.8 -15.0 10.0
Absolute poll change since last quarter
t-1
935 1.2 1.4 0.0 15.0
Government party 1072 0.4 0.5 0.0 1.0
Fáil, the Gren Party, the Progressive Democrats, Sinn Féin; for the Netherlands: CDA,
ChristenUnie, D66, GroenLinks, LPF, PvdA, PVV, SGP, SP, and VVD. The only party
that has been excluded from the analysis is the Party for the Animals (Netherlands),
because its focus on animal rights and related issues distorted the estimation of parlia-
mentary positions: because it consistently dealt with those specic issues in its manifesto
and in parliament, it was estimated to be much more left-wing than any other party.
Descriptive statistics for the variables included in the analysis are available in Table 2.
Model
The strategy for estimating the eect of opinion polls on congruence, is based on Achen's
(1978) linear regression method for measuring responsiveness of politicians to voter pref-
erences. This model is adapted to measure the predictive value of the manifesto position
mi
for the parliamentary position
pi
of a party
i
:
pi=β0+β1mi+
(1)
Achen argues that in an `unbiased' system, the intercept
β0
would equal zero and
the slope
β1
would equal 1. Our hypotheses expect the eect of the manifesto will vary
according to a party's standing in the polls
(oi)
, which is modelled as an interaction eect:
pi=β0+β1mi+β2oi+β3mioi+
(2)
When this eect, in turn, is modelled as conditional on the party's status as a gov-
ernment or opposition party (
gi)
, this results in a three-way interaction model:
pi=β0+β1mi+β2oi+β3gi+β4mioi+β5migi+β6oigi+β7mioigi+
(3)
The data analysis strategy takes into account that our observations are organised in
a multilevel fashion. Therefore, I am using a multilevel linear regression with a random
intercept for the term-quarter (Gelman and Hill, 2007). Thus for each set of parties in
a given term and quarter, the intercept of the model is allowed to vary. The discussion
8
of the eects will focus on the marginal eects, using the suggestions by Brambor, Clark
and Golder (2006).
4
I also include a lagged dependent variable in the model to correct
for autocorrelation of parliamentary party positions between quarters.
Results
Figure 1 presents manifesto and parliamentary policy positions in one (typical) term in
each country to give an idea of the amount of variation over time. In the United Kingdom
(1997-2001) we see that the manifesto ordering of parties, Liberal Democrats - Labour
- Conservative, is changed in parliament to Labour - Liberal Democrat - Conservative.
Given the fact that Labour was in government at that stage, it is not so surprising to see
that the LibDems would side more with the Conservative opposition, as is indeed seen
in parliamentary voting behaviour (Spirling and McLean, 2007). We see that over time,
the Liberal Democrats move towards the Conservatives, while Labour MPs also seem to
drift towards the centre after the initial left-wing move.
In the Irish case we see another example of the impact of government-opposition
patterns on parliamentary position-taking. Fianna Fáil, which had been in government
since 1997 coalesced with the Green Party from 2007 onwards. For the Green Party this
seems to imply an immediate shift towards their coalition party: they seemed to change
their use of words signicantly as a result of entering the coalition. Fine Gael, the largest
opposition party, however, shifts from an ideological position almost identical to Fianna
Fáil to a more centrist one. Labour and Sinn Féin seem relatively consistent in their
policy position.
For the Netherlands the picture is more garbled with nine parties included in the
2003-2006 parliament. All in all, we see the emergence of three blocks: the government
parties (CDA, VVD and D66), the left wing opposition (PvdA, SP and GL) and the
right wing opposition (SGP, LPF and CU). One should not over-emphasize some of the
changes in parties' positions, such as the strong shift of GreenLeft to the centre and then
back, which does not seem to accurately reect the parties' position-taking. Likewise, the
smaller Christian parties (CU and SGP) seem to be located quite far to the right, while
on many issues they are more centrist. This might be the result of their distinctive use
of biblical terms as well as their issue ownership of medical-ethical issues on which they
take very conservative positions (also see Louwerse, 2011
a
, 220). If these factors aect
our estimation of party policy positions in parliament, at least this is done so similarly
across a parliamentary term.
How polls aect congruence
Table 3 provides the outcome of regression analyses that test the Losers Drift Hypothesis
(Models 1 and 3) and the Poll Change Hypothesis (Models 2 and 4). These models do
not dierentiate between government and opposition parties. As three-way interaction
4
The code for the interaction plots is based on Strezhnev (2013).
9
Figure 1: Manifesto and parliamentary positions
(a) United Kingdom, 1997-2001
Parliament − Quarter
Position
Conservative
Conservative
●
●
●
●
●●●●
●
●
●●
●Labour
Labour
LibDem
LibDem
Manifesto 3 4 5 7 8 9 11 12 13 15 16
34567
(b) Ireland, 2007-2011
Parliament − Quarter
Position
FF
FF ●
●
●
●
●
●
●●
●
●
●
●
●
●●FG
FG GP
GP LAB
LAB
●
●
●
●
●
●
●
●
●
●
●●
●
●
●SF
SF
Manifesto 2 3 4 5 6 7 8 9 10 11 12 13 14 15
0246
(c) The Netherlands, 2003-2006
Parliament − Quarter
Position
CDA
CDA
●
●●●
●
●
●
●
●●
●
●
●
●
●
●CU
CU
D66
D66
GL
GL
●
●
●
●
●●●
●●
●
●
●
●
●
●
●LPF
LPF
PvdA
PvdA
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●SGP
SGP
SP
SP
VVD
VVD
Manifesto 2 3 4 5 6 7 8 9 10 11 12 13 14 15
0246810
10
Table 3: Multilevel linear regression models of parliamentary policy positions
Model 1 Model 2 Model 3 Model 4
(Intercept)
0.24∗0.32∗0.24∗0.17
(0.10) (0.13) (0.09) (0.12)
Parliament position
t-1
0.68∗∗∗ 0.68∗∗∗ 0.68∗∗∗ 0.68∗∗∗
(0.02) (0.02) (0.02) (0.02)
Manifesto position
0.26∗∗∗ 0.26∗∗∗ 0.26∗∗∗ 0.29∗∗∗
(0.03) (0.03) (0.03) (0.03)
Poll change since election
t-1
0.01
(0.02)
Manifesto * Poll change
t-1
−0.00
(0.00)
Absolute poll change since election
t-1
−0.02
(0.03)
Manifesto * Abs. poll change
t-1
0.00
(0.01)
Poll change since last quarter
t-1
0.12∗
(0.06)
Manifesto * Poll ch. lst qrt
t-1
−0.02∗
(0.01)
Absolute poll change since last quarter
t-1
0.07
(0.08)
Manifesto * Abs. poll ch. lst qrt
t-1
−0.02
(0.01)
AIC 2610.77 2609.73 2596.04 2593.28
BIC 2644.68 2643.64 2629.92 2627.17
Log Likelihood -1298.38 -1297.86 -1291.02 -1289.64
Num. obs. 939 939 935 935
Num. groups: term:quarter 177 177 177 177
Variance: term:quarter.(Intercept) 0.16 0.16 0.16 0.16
Variance: Residual 0.79 0.79 0.79 0.79
∗∗∗p < 0.001
,
∗∗p < 0.01
,
∗p < 0.05
eects are dicult to interpret from the regression table, these eects are visualised in
Figure 2.
Model 1 tests the Losers Drift Hypothesis operationalising opinion poll gains or losses
as the dierence between the election result and the (average of) current opinion polls
(at t - 1). While both the lagged dependent variable as well as the manifesto position
are signicantly related to the current parliamentary policy position, I do not nd that
poll gains or losses signicantly alter the marginal eect of the manifesto position on the
parliamentary position. Model 2 tests the Poll Change Hypothesis and similarly nds no
eect of parties absolute change in the polls on the congruence between manifesto and
parliamentary positions. When we look at parties' change in polls
compared to the last
quarter
, we do nd a signicant eect (Model 3). The marginal eect of manifesto scores
is about 0.55 for the worst loss in the polls, while it is statistically indistinguishable from
11
Figure 2: Marginal eects of manifesto position on parliamentary position (Table 3)
(a) Model 1
−20 −10 0 10
0.1 0.2 0.3 0.4 0.5
Change in party polls (t−1)
Estimated marginal coefficient
(b) Model 2
0 5 10 15 20 25
0.1 0.2 0.3 0.4 0.5
Absolute change in party polls (t−1)
Estimated marginal coefficient
(c) Model 3
−15 −10 −5 0 5 10
0.0 0.2 0.4 0.6 0.8
Change in party polls, quarter−on−quarter (t−1)
Estimated marginal coefficient
(d) Model 4
0 5 10 15
−0.4 −0.2 0.0 0.2
Absolute change in party polls, quarter−on−quarter (t−1)
Estimated marginal coefficient
zero when looking at the biggest quarter on quarter improvement in the polls. This eect
is not statistically signicant, however, if we delete ve extreme cases with poll gains or
losses over 7.5%. Therefore Model 3 at best provides tentative support for the Losers
Drift Hypothesis. Model 4 does not nd a signicant interaction eect between manifesto
support and the absolute change in poll numbers since the previous quarter.
One explanation for the ndings in Table 3 might be that the eects of opinion polls are
dierent for government and opposition parties and therefore cancel each other out. These
conditional hypotheses are tested in Table 4 and the interaction eects are visually dis-
played in Figure 3. Model 1 suggests that change in party polls since the election does not
have a dierential eect on government and opposition parties' manifesto-parliamentary
position congruence. Model 3 provides the same result when looking at quarterly change
in the polls. Therefore, the analysis does not show support for the Conditional Losers
Drift Hypothesis.
Models 2 and 4 show support for the Conditional Poll Change Hypothesis: government
parties that see a lot of change in their polling numbers generally show a higher marginal
eect of manifesto positions on parliamentary positions, but this is not true for opposition
parties - if anything they show the reverse. This holds both for the change in poll numbers
since the last election as well as changes in poll numbers compared to the previous quarter.
12
Table 4: Multilevel linear regression models of parliamentary policy positions (with gov-
ernment interaction eect)
Model 1 Model 2 Model 3 Model 4
(Intercept)
0.19 0.20 0.17 0.08
(0.11) (0.13) (0.10) (0.13)
Parliament position
t-1
0.68∗∗∗ 0.66∗∗∗ 0.68∗∗∗ 0.66∗∗∗
(0.02) (0.02) (0.02) (0.02)
Manifesto position
0.28∗∗∗ 0.32∗∗∗ 0.28∗∗∗ 0.34∗∗∗
(0.03) (0.03) (0.03) (0.03)
Government party
0.63∗1.00∗∗ 0.52 0.91∗
(0.29) (0.35) (0.27) (0.37)
Manifesto * Government
−0.11∗−0.20∗∗ −0.09 −0.18∗∗
(0.05) (0.06) (0.05) (0.06)
Poll change since election
t-1
0.01
(0.03)
Manifesto * Poll change
t-1
−0.00
(0.00)
Poll change
t-1
* Government
0.10
(0.06)
Manifesto * Poll change
t-1
* Gov.
−0.01
(0.01)
Absolute poll change since election
t-1
0.00
(0.03)
Manifesto * Abs. poll change
t-1
−0.01
(0.01)
Abs. poll change
t-1
* Government
−0.18∗
(0.08)
Manifesto * Abs. poll change
t-1
* Gov.
0.04∗∗
(0.01)
Poll change since last quarter
t-1
0.12
(0.07)
Manifesto * Poll ch. lst qrt
t-1
−0.02
(0.01)
Poll ch. lst qrt
t-1
* Government
0.09
(0.15)
Man. * Poll ch. lst qrt
t-1
* Gov.
−0.01
(0.03)
Absolute poll change since last quarter
t-1
0.14
(0.09)
Manifesto * Abs. poll ch. lst qrt
t-1
−0.05∗∗
(0.02)
Abs. poll ch. lst qrt
t-1
* Government
−0.43∗
(0.21)
Man. * Abs. poll ch. lst qrt
t-1
* Gov.
0.10∗∗
(0.03)
AIC 2634.44 2621.49 2618.32 2601.62
BIC 2687.73 2674.78 2671.57 2654.87
Log Likelihood -1306.22 -1299.74 -1298.16 -1289.81
Num. obs. 939 939 935 935
Num. groups: term:quarter 177 177 177 177
Variance: term:quarter.(Intercept) 0.16 0.16 0.16 0.16
Variance: Residual 0.79 0.78 0.79 0.78
∗∗∗p < 0.001
,
∗∗p < 0.01
,
∗p < 0.05
13
Figure 3: Marginal eects of manifesto position on parliamentary position (Table 4)
(a) Model 1
−20 −10 0 10
−0.5 0.0 0.5 1.0
Change in party polls (t−1)
Estimated marginal coefficient
Opposition
Government
(b) Model 2
0 5 10 15 20 25
0.0 0.5 1.0
Absolute change in party polls (t−1)
Estimated marginal coefficient
Opposition
Government
(c) Model 3
−15 −10 −5 0 5 10
−0.5 0.0 0.5 1.0
Change in party polls, quarter−on−quarter (t−1)
Estimated marginal coefficient
Opposition
Government
(d) Model 4
0 5 10 15
−0.5 0.0 0.5 1.0 1.5
Absolute change in party polls, quarter−on−quarter (t−1)
Estimated marginal coefficient
Opposition
Government
The eect does not change substantively if we exclude extreme cases of changing poll
numbers.
Discussion and conclusion
The empirical results suggest that opinion poll change mediates the congruence between
manifesto and parliamentary policy positions of political parties. Whereas there are some
tentative indications that poll losses are related to a higher degree of congruence, stronger
eects are found when we look at the impact of the absolute poll change (i.e. gaining
or
losing). The data showed support for the Conditional Poll Change Hypothesis, which
states that poll changes decrease congruence for opposition parties, but increase it for
government parties.
Caution is, however, warranted in the causal interpretation of these results. Perhaps
voters adjust their voting intentions (in polls) based on how parties perform their repres-
entative role. Larger change in opinion polls might be the result rather than the cause
of policy incongruence. Most of the explanatory variables are lagged to account for this
and a lagged dependent variable is included in an attempt to exclude this alternative
14
explanation.
Another concern might be that the current analysis is limited to a small number of
countries and only one issue dimension, the left-right scale. Including a more extensive
selection of countries might be helpful in exploring cross-country dierences, which a rel-
atively low number of observation per country made dicult here. Apart from extending
the number of countries (and the time frame within countries), multiple issue dimensions
could be included in the analysis. After all, left-right is a limited summary of political
conict in many countries.
Given that our analysis suggests that both big losses or big gains in the polls aect
congruence, the most important challenge is to track parties' responses to opinion polls
more in-depth. Parties seem to respond dierently to opinion polls, depending on the
exact circumstance they nd themselves in. The current article looked at the distinction
between government and opposition parties, but other factors could also play a role, such
as the party type (catch all, single issue), party size and how well it performed at the last
election. It is therefore relevant to complement large-N analyses of congruence with case
studies to better track the exact mechanisms at work.
References
Achen, Christopher H. 1978. Measuring Representation.
American Journal of Political
Science
22(3):475510.
APSR. 1950. Toward a More Responsible Two-Party System: A report of the Committee
on Political Parties.
American Political Science Review
44(3, Part 2: Supplement).
Bakker, Ryan, Catherine De Vries, Erica Edwards, Liesbeth Hooghe, Seth Jolly, Gary
Marks, Jonathan Polk, Jan Rovny, Marco Steenbergen and Milada Anna Vachudova.
2012. Measuring party positions in Europe: The Chapel Hill expert survey trend le,
1999-2010.
Party Politics
.
URL:
http://ppq.sagepub.com/cgi/doi/10.1177/1354068812462931
Benoit, Kenneth, Thomas Bräuninger and Marc Debus. 2009. Challenges for Estimat-
ing Policy Preferences: Announcing an Open Access Archive of Political Documents.
German Politics
18(3):441454.
Brambor, T, W R Clark and M Golder. 2006. Understanding interaction models: Im-
proving empirical analyses.
Political Analysis
14(1):6382.
Budge, Ian, Judith Bara, Andrea Volkens and Hans-Dieter Klingemann. 2001.
Mapping
policy preferences : estimates for parties, electors and governments, 1945-98
. Oxford:
Oxford University Press.
Dalton, Russell J., David M. Farrell and Ian McAllister. 2011.
Political parties and
democratic linkage : how parties organize democracy
. Oxford; New York: Oxford
University Press.
15
Fivaz, Jan, Tom Louwerse and Daniel Schwarz. 2014. Keeping Promises: Voting Advice
Applications and Political Representation. In
Matching Voters with Parties and Can-
didates: Voting Advice Applications in Comparative Perspective
, ed. Diego Garzia and
Stefan Marschall. Colchester: ECPR Press pp. 197 216.
URL:
http://press.ecpr.eu/book_details.asp?bookTitleID=104
Gelman, Andrew and Jennifer Hill. 2007.
Data analysis using regression and multi-
level/hierarchical models. Analytical Methods for Social Research
. Cambridge: Cam-
bridge University Press.
Golder, Matt and Jacek Stramski. 2010. Ideological Congruence and Electoral Institu-
tions.
American Journal of Political Science
54(1):90106.
URL:
http://dx.doi.org/10.1111/j.1540-5907.2009.00420.x
Guardian Data. 2014. All Guardian/ICM poll results..
URL:
http://goo.gl/LDIK9H
Hobolt, Sara Binzer and Robert Klemmensen. 2008. Government Responsiveness
and Political Competition in Comparative Perspective.
Comparative Political Stud-
ies
41(3):309337.
URL:
http://cps.sagepub.com/cgi/doi/10.1177/0010414006297169
Klingemann, Hans-Dieter, Richard I. Hoerbert and Ian Budge. 1994.
Parties, policies,
and democracy
. Theoretical lenses on public policy Boulder: Westview Press.
Laver, Michael J., Kenneth Benoit and John Garry. 2003. Extracting Policy Posi-
tions from Political Texts Using Words as Data.
American Political Science Review
97(2):311331.
Lijphart. 2012.
Patterns of Democracy: Government Forms and Performance in Thirty-
Six Countries
. 2nd ed. New Haven: Yale University Press.
Louwerse, Tom. 2011
a
. Political Parties and the Democratic Mandate: Comparing Col-
lective Mandate Fullment in the United Kingdom and the Netherlands Phd thesis
Universiteit Leiden.
Louwerse, Tom. 2011
b
. The Spatial Approach to the Party Mandate.
Parliamentary
Aairs
64(3):425447.
URL:
http://pa.oxfordjournals.org/cgi/doi/10.1093/pa/gsr012
Louwerse, Tom. 2012. Mechanisms of Issue Congruence: The Democratic Party Man-
date.
West European Politics
35(6):12491271.
URL:
http://www.tandfonline.com/doi/abs/10.1080/01402382.2012.713745
Mair, Peter. 2009. Representative versus Responsible Government..
URL:
http://edoc.vifapol.de/opus/volltexte/2010/2121/
16
Mansergh, Lucy and Robert Thomson. 2007. Election pledges, party competition, and
policymaking.
Comparative Politics
39(3):311329.
URL:
http://www.jstor.org/stable/10.2307/20434043
Marks, Gary, Liesbet Hooghe, Marco R. Steenbergen and Ryan Bakker. 2007. Cross-
validating data on party positioning on European integration.
Electoral Studies
26(1):2338.
URL:
http://www.sciencedirect.com/science/article/B6V9P-4KCRS57-
1/2/ae12c09e09ca531f08a85ba21866bdc4
Marsh, Michael. 2015. Irish Opinion Poll Archive..
URL:
http://www.tcd.ie/Political_Science/IOPA/
Martin, Lanny W. and Georg Vanberg. 2007. A Robust Transformation Procedure for
Interpreting Political Text.
Political Analysis
16(1):93100.
McDonald, Michael D. and Ian Budge. 2005.
Elections, parties, democracy : conferring
the median mandate
. Oxford, New York: Oxford University Press.
Naurin, Elin. 2002. The Pledge Paradox. In
ECPR Joint Session
. Number March Turin:
pp. 2227.
Oce of the Houses of the Oireachtas. 2014. Oireachtas debates in XML format 1921-
2012..
URL:
http://debates.oireachtas.ie/XML/
Oce of the Houses of the Oireachtas. 2015. Current Debates..
URL:
http://oireachtasdebates.oireachtas.ie/
Ociële Bekendmakingen. 2012. Ociële Bekendmakingen..
URL:
https://zoek.ocielebekendmakingen.nl/
Parlementair Documentatie Centrum. 2010. Parlement & Politiek.
\
url{http://www.parlement.com}.
Parliamentary Parser Project. 2015. Hansard Reports..
URL:
http://parser.theyworkforyou.com/hansard.html
Pellikaan, Huib, Gerrit Voerman and Joop J.M. van Holsteyn. 2003.
Verkiezinsgpro-
gramma's: verkiezingen van de Tweede Kamer 15 mei 2002 & 22 januari 2003
. Ams-
terdam: Rozenberg Publishers.
Pellikaan, Huib, Joop J.M. van Holsteyn, T der Meer, H IJsbrandy and Gerrit Voerman.
2006.
Verkiezingen van de Tweede Kamer 22 november 2006: Verkiezingsprogramma's
met CD-ROM
. Amsterdam: Rozenberg Publishers.
Powell, G. Bingham. 2000.
Elections as Instruments of Democracy
. New Haven: Yale
University Press.
17
Ranney, A. 1954.
The Doctrine of Responsible Party Government
. 1962 repr. ed. Urbana
Illinois: University of Illinois Press.
Royed, Terry J. 1996. Testing the Mandate Model in Britain and the United States:
Evidence from the Reagan and Thatcher Eras.
British Journal of Political Science
26(1):45.
URL:
http://www.journals.cambridge.org/abstract_S0007123400007419
Seawright, Jason and John Gerring. 2008. Case Selection Techniques in Case Study Re-
search: A Menu of Qualitative and Quantitative Options.
Political Research Quarterly
61(2):294308.
URL:
http://prq.sagepub.com/cgi/doi/10.1177/1065912907313077
Somer-Topcu, Zeynep. 2009. Timely Decisions: The Eects of Past National Elections
on Party Policy Change.
The Journal of Politics
71(1):238.
Soroka, Stuart Neil and Christopher Wlezien. 2010.
Degrees of democracy politics, public
opinion, and policy
. New York: Cambridge University Press.
URL:
http://site.ebrary.com/id/10360055
Spirling, A. and I. McLean. 2007. UK OC OK? Interpreting Optimal Classication
Scores for the U.K. House of Commons.
Political Analysis
15(1):8596.
URL:
http://pan.oxfordjournals.org/cgi/doi/10.1093/pan/mpl009
Stimson, James A., Michael B. Mackuen and Robert S. Erikson. 1995. Dynamic Rep-
resentation.
American Political Science Review
89(3):543.
Strezhnev, Anton. 2013. Marginal Eect Plots for Interaction Models in R..
URL:
http://causalloop.blogspot.ie/2013/06/marginal-eect-plots-for-interaction.html
Strøm, Kaare. 1990. A Behavioral Theory of Competitive Political Parties.
American
Journal of Political Science
34(2):565598.
URL:
http://www.jstor.org/stable/2111461
Strøm, Kaare. 1994. The Presthus Debacle: Intraparty Politics and Bargaining Failure
in Norway.
The American Political Science Review
88(1):112.
URL:
http://www.jstor.org/stable/2944885?origin=crossref
Thomassen, Jacques J.A. 1994. Empirical Research into Political Representation: Failing
Democracy or Failing Models? In
Elections at Home and Abroad
, ed. M K Jennings
and T E Mann. Ann Arbor: University of Michigan Press pp. 237265.
Thomson, Robert. 1999. The Party Mandate. Election Pledges and Government Actions
in the Netherlands, 1986-1998 PhD thesis RUG / UU / KUN: ICS.
Thomson, Robert. 2001. The programme to policy linkage: The fullment of election
pledges on socio-economic policy in the Netherlands, 1986-1998.
European Journal of
Political Research
40(2):171197.
18
Thomson, Robert, Terry J Royed, Elin Naurin, Joaquín Artés, Marc Ferguson, Petia
Kostadinova and Catherine Moury. 2012. The program-to-policy linkage: a compar-
ative study of election pledges and government policies in ten countries..
Warwick, Paul V. 2014. Public opinion and government policy in Britain : A case of
congruence , amplication or dampening ?
European Journal of Political Research
54(1):n/an/a.
URL:
http://doi.wiley.com/10.1111/1475-6765.12069
Wlezien, Christopher and Stuart Neil Soroka. 2012. Political Institutions and the
Opinion-Policy Link.
West European Politics
35(6):14071432.
URL:
http://www.tandfonline.com/doi/abs/10.1080/01402382.2012.713752
19
