Upload
lycong
View
218
Download
0
Embed Size (px)
Citation preview
Is Social Left/Right Cross-Nationally Comparable inEuropean Parties?
Ryan Bakker, University of GeorgiaSeth Jolly, Syracuse University
Jonathan Polk, University of Gothenburg
April 4, 2016
Abstract
Using survey vignettes and scaling techniques, we estimate a common social left/rightdimension for political parties across the member states of the European Union. Pre-vious research shows that economic left/right travels well across the EU, meaning thatthe placements of parties on that dimension are cross-nationally comparable; how-ever, the social dimension is more complex, with different issues forming the core ofthe social dimension in different countries. The 2014 wave of the Chapel Hill ExpertSurvey includes anchoring vignettes which we use as “bridge votes” to place partiesfrom different countries on a common social left/right dimension. We estimate thedimension using the “blackbox” technique. The resulting scale offers a cross-nationallycomparable interval-level measure of a party’s social left/right ideological position.
Expert surveys are widely used to obtain information on the policy positions of political
parties in contemporary democracies (Ray 1999; Benoit and Laver 2006; Hooghe et al. 2010;
Rohrschneider and Whitefield 2012; Bakker et al. 2015). One concern about the expert
survey approach to measuring party positions pertains to the cross-national comparability of
respondent placements (Budge 2000; McDonald, Mendes, and Kim 2007). If experts for each
country place only the parties within a single party system, can we be sure that respondents
for France conceive of and use a particular scale in the same way as their colleagues that
complete a survey for the parties of Lithuania?
In previous work, we combined ‘blackbox’ scaling techniques (Poole 1998) with a series of an-
choring vignettes (King et al. 2004; King and Wand 2007) embedded within the 2010 Chapel
Hill Expert Survey to show that expert placements of political parties on the economic left-
right dimension are cross-nationally comparable across Europe (Bakker et al. 2014). Yet, we
also know that politics in many contemporary European democracies is multidimensional,
and that a “second” dimension is an important determinant of public opinion and party com-
petition on issues such as immigration and European integration (Hooghe and Marks 2009;
De Vries and Hobolt 2012; Rovny 2014; Hausermann and Kriesi 2015; Hobolt and de Vries
2015). As challenger parties that emphasize this cultural dimension become more prominent
across Europe, we require a deeper understanding of the cross-national comparability of this
more complex dimension in today’s politics.
In addition to the social dimension, European integration has taken on increasing political
importance, as the bailouts brought on by the financial crisis and contemporary migration
challenges make clear. And although prominent theorists of representation have argued that
the quality of representation at the European level is actually rather high, ‘the empirical
evidence...in support of this claim refers only to the left-right dimension which, while dom-
inant, does not constrain all of the policy preferences of voters and political parties” (Mair
and Thomassen 2010, 30). In short, while there is growing evidence for the cross-national
1
comparability of the left-right dimension for both expert survey respondents and European
politics more broadly, it remains unclear if this is the case for other prominent dimensions
of political competition in contemporary European societies.
This paper addresses questions of cross-national comparability for social left-right and Eu-
ropean integration by extending on the combination of anchoring vignettes embedded in the
2014 Chapel Hill Expert Survey (CHES) on party positions in Europe and Bayesian scaling
techniques with foundations in the work of Aldrich and McKelvey (Hare et al. 2015). We
begin by briefly describing the 2014 wave of the CHES data, with a focus on the anchoring
vignettes pertaining to economic left-right, social left-right, and European integration posi-
tions. We then provide an overview of the Bayesian Aldrich-McKelvey scaling procedures
that are used to recover cross-nationally comparable scaled positions on these three dimen-
sions for political parties in Europe. These scaled dimensions are then used to replicate a
prominent study on the relationship between economic and social left-right positions, and
party positions on European integration (Marks et al. 2006). We report that the central
findings of that article hold when using the scaled data from 12 years later.
1 CHES
The 2014 Chapel Hill Expert Survey on party positions in Europe is the most recent wave in
an ongoing research project designed to measure the positions of political party leadership
on dimensions and policies related to the economy, socio-cultural matters, and European
integration (Hooghe et al. 2010; Bakker et al. 2015). 337 political scientists that study
political parties and/or European integration completed the survey, and all 28 EU members,
plus Norway, Switzerland, and Turkey, were included in the 2014 round. This produced
information for a total of 268 political parties in Europe.
Each expert respondent placed the parties of only one party system, which creates some
2
uncertainty as to whether or not the expert respondents for one country differ systematically
in their use of the various policy scales from experts that completed the survey for a different
country. In order to address this concern, a series of anchoring vignettes were presented to
every respondent at the end of the survey (King et al. 2004; King and Wand 2007; Bakker
et al. 2014). This vignette section described the positions of three hypothetical parties for
three dimensions: economic left-right, social left-right, and European integration.1 Unlike
the majority of the survey, every CHES respondent had the opportunity to place these
hypothetical vignette parties on the same three scales, and these placements can then be
used as bridging information to facilitate comparison of the respondents’ other placements.
The next section describes this process in more detail and presents the scaled positions for
the parties included in the 2014 CHES data.
2 CHES meets BAM!
When survey respondents in different contexts answer the same survey questions, there is
the potential that these respondents have different interpretations of the response categories.
This is especially true when the response categories represent relative positions on a latent
scale, such as left-right ideology. Differential-item functioning (DIF) occurs when such an
underlying scale is interpreted differently across a range of respondents. DIF can lead to a
distortion of the placements of stimuli (i.e. political parties) on a given scale. For example,
a British party expert and a Greek party expert may view the end points of a pro/anti
EU integration scale in slightly different ways, confounding the ability to compare parties’
positions across these two countries. Although this is not an important distinction within a
given country, these potential disortations could be problematic when comparing positions
of parties across countries. In the late 1970s, Aldrich and McKelvey (1977) developed an
estimation technique aimed at correcting DIF in perception/placement scales. Their solution,
1The wording of the vignettes can be found below in Appendix A.
3
Aldrich-McKelvey scaling (henceforth A-M), assumes that there exists a true placement for
a given stimuli (i.e. candidate or party) and that any individual placement of a stimuli
is actually a linear distortion of this true placement. The A-M solution, then, allows each
survey respondent to have her own ‘distortion’ parameters, while treating the true placement
of a stimuli as fixed across all respondents. The distortion parameters act in such a way
as to allow different respondents to shift the true placement to the left or right and/or to
expand or contract distances between placements on the underlying scale.
A-M takes advantage of the fact that survey respondents in their data all place the same set
of stimuli as well as placing themselves on an ideological left-right scale. This combination
of information allows for estimation of the individual distortion parameters and results in a
scale in which the placements of the repsondents and stimuli are meaningfully comparable.
While Aldrich and McKelvey were interested in comparing ideological self-placements of
survey respondents and a set of political stimuli, the technique they developed also allows for
comparison of political stimuli across different contexts. We take advantage of their insight
and employ a Bayesian version of this estimator to compare placements of political parties
in different countries, where the placements are derived from country-specific respondents.
That is, rather than comparing the ideological self-placement of voters to their perceptions
of the location of candidates on the same scale, we are interested in the relative placement
of political parties that were placed by different experts on the same scale. As long as all of
the respondents place at least one common stimuli, the solution to this problem is relatively
straightforward.
As much of an innovation as A-M scaling represents in terms of overcoming problems with
DIF, there are some serious limitations. Most notably for our purposes, A-M scaling does
not allow for missing data when deriving a cross-contextually comparable scale. Given the
nature of the CHES data, this is seriously problematic. The CHES asks country-specific
experts to place parties on a variety of dimensions, but only in the country in which that
4
respondent is an expert. That is, British experts place only British parties and Greek experts
place only Greek parties. When all of these country-specific data are combined, the resulting
data matrix contains huge amounts of missing data, as British experts do not place Greek
parties, etc. With this structure to our data, classic A-M scaling is not an option. A second,
less problematic limitation is that classic A-M scaling does not yield estimates of uncertainty
for the estimated stimuli positions. This limitation makes it impossible to discern whether
or not two different candidates/parties are statistically distinguishable from one another on
some dimension. For example, we would not be able to statistically determine whether or
not one party was more pro-EU than another party. While there are options to approximate
uncertainty estimates (i.e., bootstrapping) the scaling procedure itself does not yield such
information.
In order to overcome these limitations, we use the Bayesian Aldrich-McKelvey (BAM) scaling
procedure developed in Hare et al. (2015). In that paper, the authors place senators, candi-
dates for the Senate, and survey respondents on a common ideological scale. Their data are
strikingly similar to ours in that only respondents from a senator’s/candidate’s home state
place that senator/candidate, which leads to an abundance of missingness in the final data
set (compiling data from all 50 states). The survey used by Hare et al. (2015) also asked
respondents to place President Obama, the Democratic and Republican parties, and the Tea
Party on the same ideological scale that they were placing their state-specific Senate candi-
dates. So, while respondents in two different states placed different senators/candidates from
one another, all respondents placed the president and the parties. These common stimuli
act as ‘bridging’ votes and are the key to developing a cross-contextually comparable scale.
As described above, the 2014 wave of the CHES includes anchoring vignettes that we use as
bridge votes to construct a cross-nationally comparable scale. The Bayesian implementation
of A-M scaling easily handles the missing data in our model as missing values are auto-
matically imputed via Markov chain Monte Carlo (MCMC). MCMC also directly produces
5
measures of uncertainty that reflect both variance in the observed placements of parties as
well as the degree of missingness for a given party. This yields larger standard errors for
parties with fewer observed placements.
The BAM model closely resembles a Bayesian factor model, with the primary distinction
being how the parameters and the latent variable are indexed. That is, the factor model
assumes that there is some latent variable, X, that is specific to a given respondent and that
this latent variable is related to observable indicators. The latent variable is related to these
observable indicators through parameters (often called factor loadings) that are indexed by
the observed indicators. In the BAM model, the indexing is reversed relative to the factor
model. This means that the the latent variable X is now indexed by observable indicator and
the parameters in the model are now indexed by respondent. For our data, this translates to
the position of a party on a specific dimension Xj, where j indexes party, being related to an
expert’s placement of that party yij, where i indexes expert, through paramters αi and βi.
Following the above discussion α and β are the distortion parameters that map the expert’s
perception of a party’s placement onto the ‘true’ position of that party.
To be Bayesian, we must specify distributional assumptions for the unknown quantities in
the model. We must also specify the distrubtion of the dependent variables, in this case the
expert placements of the parties, yij. As these placements are 11 point scales, we assume
them to be normally distributed with an estimated mean and variance. We then set the
mean position of expert i’s placement of party j to be equal to the the true position of party
j, Xj, as well as the parameters αi and βi. Formally, the model is:
yij ∼ N(µij, τij)
µij = αi + βiXj
The unknown quantities, α, β, and X require prior distributions in the Bayesian setup. For
this model, we specify uniform priors for α and β and a standard Normal prior for X. Finally,
6
we specify an uniformed conjugate Gamma prior for the variance of expert placements, τij.
Formally:
αi ∼ Uniform(−100, 100)
βi ∼ Uniform(0, 100)
Xj ∼ N(0, 1)
τij ∼ Gamma(1, .1)
In order to identify the model and to set the scale of the latent variable, we constrain the βs
to be positive. This assures that higher values of the latent variable, X, are associated with
higher values of the expert placements. Substantively, this means for the economic and social
left-right dimensions, higher values of the latent variable indicate more right-wing positions
whereas for the EU dimension, higher values of the latent variable represent a more pro-
EU position. As an additional identification constraint, we specified prior positions for the
vignette party placements that respects the intended ordering of the vignette placements.
This is what King et al. (2004) refer to as vignette equivalence and is a requirement of
anchoring vignette-based scaling. That is, in order to be included in the model, experts
must correctly perceive the ordering of the vignette parties. We require that each party be
placed by at least 3 experts in order to be included in the estimation. With these restrictions,
we are able to produce a cross-nationally comparable scale for the economic left-right, social
left-right, and pro/anti-EU integraton dimension for 259 parties based on the input of 333
experts. We estimated the model using JAGS (Just Another Gibbs Sampler) via the R
package rjags. For each dimension, we ran two chains for 20, 000 iterations, discarding the
first 5, 000 as a burn-in. The chains show strong evidence of convergence across a variety of
diagnostics.
After running the BAM procedure for each of the three dimensions for which we have vignette
placements, we then sample 1, 000 draws from the posterior distribution of each party’s
7
placement on each dimension. We then summarize these distributions grapically by plotting
the mean and 95% interval of each party’s distribution for each dimension. We present these
distributions in the following graphs, but separate each dimension into two graphs in order to
make reading the y-axis labels easier. Figures 1 and 2 plot the left-leaning and right-leaning
parties in terms of economic positions while Figures 3 and 4 present the same on the social
dimension. Finally, Figures 5 and 6 display the distributions of parties on the European
dimension.
As Figures 1–6 illustrate, there is a great deal of variation across these parties in terms of
their positions on these three dimenssions. These plots help us to identify the most extreme
parties on these various dimensions across the members of the EU. For example, the Greek
KKE party is the most left-wing in terms of economic left-right position whereas the the
Slovenian SDS party is the most economically right-wing party. Similarly, the most left-wing
parties in terms of social policy are the green parties of Belgium and the UK while the most
right-wing in terms of the social dimension are the German NPD party and Britain’s UKIP.
Finally, in terms of the EU dimension, the liberal parties of Italy and Finland are at the
extreme pro-EU end of the scale whereas the Greek KKE and the UK’s UKIP have the most
anti-EU integration positions. These results pass the face validity test, as the parties line up
where we would expect them to. It is also worth noting that the vignette parties, labeled
as A,B, and C are not the most extreme parties on any dimension but, rather, the other
parties are placed relative to the experts’ perception of where the vignette parties fall.
8
Party Placements with 95% Credible Intervals
Economic Left−Right: left parties
gre_KKEbel_PVDA
sle_ZLpor_CDU
den_ELpor_BE
it_RCgre_SYRIZA
ire_SPire_PBPA
it_SELden_FolkBcze_KSCM
ger_Linkelux_DL
spa_Amaiurspa_Podemos
fra_PGnet_SP
tur_HDPbel_Groen
den_SFswe_V
uk_GREENbel_ECOLO
_Bspa_IUswe_FI
rom_PP−DDbel_PS
bul_ATAKAlat_LKS
spa_ICVire_SF
fin_VASspa_BNG
swi_GPS/PESger_DieTier
aus_GRUNEaus_SPO
cro_HL−SRbel_SPA
swi_SP/PSslo_Smer−SD
sle_SDcze_CSSDnet_PvdDuk_PLAID
fra_Enscyp_AKELlat_SDPS
gre_XAuk_SNP
gre_DIMARsle_PSnet_GLpol_PiS
lith_LSDPger_Piratenger_Grunen
fra_EELVrom_UNPR
den_SDsle_DeSUS
hun_LMPger_SPD
lith_DKpol_SLDtur_CHP
lith_LVZShun_JOBBIK
swe_SAPlith_LLRA
fin_SDPpol_SP
hun_Fideszswe_MPpol_PSLcze_SZ
net_PvdAcro_ORaHspa_ERCrom_PSD
lat_NSLit_M5S
bul_ABVire_GP
uk_LABbul_VMRO−BND
net_50PLUSest_EK
hun_MSZPfin_PS
bul_BSPbel_cdHfra_PRGire_Lab
sle_ZaABfra_PS
swi_CSP/PCSden_DF
fin_VIHRpol_RP
spa_PSOEslo_SNS
por_PSnet_CU
bul_NFSBger_NPD
lith_TTspa_EA
bul_BBTlat_LRA
cro_HSScyp_EDEK
lith_DPit_PD
_Aest_SDE
lux_GRENGcze_USVIT
rom_PCbul_DPScyp_KOP
lat_ZZSnet_PVV
it_VdAswe_PIRAT
fra_MODEMswi_EVP/PEV
fin_KESK
−1.5 −1.0 −0.5 0.0 0.5 1.0
●●
●●
●●
●●
●●●●●●●
●●
●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●
Figure 1: Economic left-right
9
Party Placements with 95% Credible Intervals
Economic Left−Right: right parties
fin_KDhun_DK
uk_LIBDEMcro_HSP−AS
bel_VBcro_SDPswe_SD
cro_HSPlux_LSAPlux_CSVlat_N_A
hun_E14sle_SMCswi_LdT
it_CDgre_ANEL
cze_KDU−CSLgre_Potamigre_PASOK
aus_FPObel_FDF
ire_FFcro_HDSSB
fra_FNcro_IDSit_UDC
slo_SMK−MKPgre_LAOS
lat_Vger_CDUcro_HDZger_CSU
it_SVPslo_KDH
bel_CD&Vit_Fdl
est_EERswi_CVP/PVC
tur_MHPcyp_DIKO
cze_ANO2011por_MPT
rom_UDMRslo_MH
aus_OVPswi_EDU/UDF
mal_PLslo_OLaNO
den_RVcro_HNSrom_PLR
it_NCDnet_CDA
spa_UpyDpol_PO
sle_SLSnet_D66
lith_TS−LKDfra_PRV
spa_EAJ/PNVspa_CC
swi_BDPpor_PP
fin_RKP/SFPpol_PR
bul_DBGslo_Siet
slo_NOVAire_FG
spa_C'srom_PNL
mal_PNnet_SGP
swi_GLP/PVLfra_AC
swe_KDest_EVE
bul_GERBaus_BZO
den_Vswe_FPfra_NC
spa_CiUrom_PDL
slo_SDKU−DSswe_M
cro_HSLSswe_C
rom_PMPgre_ND
it_FIaus_NEOS
fin_KOKcze_TOP09
fra_UMPpor_PSD
aus_TeamStronachlux_ADRbul_SDS
it_LNden_KF
swi_SVP/UDCger_FDP
cyp_DISYcyp_EVROKOswi_FDP/PLR
cze_ODSit_SC
tur_AKPspa_PPest_ER
uk_CONSbul_DSBger_AfD
net_VVDbel_MRest_IRL
fra_MPFbel_VLDslo_SaSlux_DP
_Cuk_UKIP
bel_N−VAlith_LRLS
bel_PPcze_SVOBODNI
den_LApol_KNPsle_NSI
sle_SDS
−1.5 −1.0 −0.5 0.0 0.5 1.0
●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●
●●
●●●●
●●
●●
●
Figure 2: Economic left-right
10
Party Placements with 95% Credible Intervals
Social Left−Right: left parties
bel_ECOLObel_Groen
it_SELuk_GREEN
fin_VIHR_A
swe_FIger_Piraten
fra_EELVsle_ZLlux_DL
lux_GRENGlux_DP
spa_ICVnet_GLpol_RP
ire_PBPAnet_D66
swe_PIRATswi_GPS/PES
den_RVcro_ORaH
swe_MPit_RC
ire_SPspa_IUpor_BE
ger_Grunenden_ELcze_SZswe_V
aus_GRUNElux_LSAP
fin_VASbel_VLD
spa_Podemosspa_BNG
fin_RKP/SFPden_LA
bel_PVDAswi_SP/PS
cro_IDScro_HNS
ire_GPbel_SPA
gre_Potamiest_SDE
mal_PLgre_SYRIZAspa_Amaiuruk_LIBDEM
tur_HDPcro_HL−SRspa_PSOE
spa_ERClith_LRLSnet_PvdD
gre_DIMARit_M5S
den_SFbel_MR
cro_SDPhun_E14ger_FDP
fra_PGhun_DKslo_SaS
it_PDaus_NEOS
fin_SDPest_ERire_Labswe_C
sle_ZaABswi_GLP/PVL
sle_PSpol_SLD
bel_PSswe_FP
hun_LMPfra_PS
fra_Ensnet_PvdA
spa_UpyDsle_SD
spa_C'sden_FolkB
bul_SDSbul_DBG
swe_SAPcro_HSLS
est_EVEuk_LAB
est_EERfra_PRG
aus_SPObul_DSB
lat_Vbel_FDFpor_PS
gre_PASOKcyp_AKEL
tur_CHPger_SPD
net_SPuk_PLAIDlith_LSDP
hun_MSZP_B
fin_KOKcze_CSSD
sle_SMCcze_ANO2011
swe_Mcyp_KOPpor_CDUger_Linke
ger_DieTiercze_SVOBODNI
net_50PLUSspa_EA
ire_SFcyp_EDEK
rom_PLRmal_PNuk_SNP
rom_PMPbel_cdH
bul_GERBpol_PO
swi_FDP/PLR
−1.5 −1.0 −0.5 0.0 0.5 1.0
●●●
●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●
Figure 3: Social left-right
11
Party Placements with 95% Credible Intervals
Social Left−Right: right parties
rom_PNLnet_VVD
sle_DeSUSden_SDbul_ABV
swi_CSP/PCSfra_MODEM
rom_PDLspa_CC
slo_SDKU−DSrom_UDMR
gre_KKEden_V
bel_CD&Vcze_TOP09
fra_PRVit_SC
ire_FGcyp_DISY
bul_BSPslo_MHlith_DP
cze_KSCMbel_N−VAcyp_DIKO
it_VdAbul_DPS
cze_ODSspa_CiU
est_EKlith_LVZSger_CDU
fra_NCfra_AC
spa_EAJ/PNVpor_PSD
uk_CONSfin_KESKnet_CDA
swi_CVP/PVCire_FF
cyp_EVROKOpor_MPT
est_IRLswe_KD
it_CDaus_TeamStronach
lat_SDPSswi_BDP
gre_NDpol_PSLslo_Siet
slo_Smer−SDlat_ZZS
slo_NOVAcze_USVIT
bul_BBTsle_SLS
rom_PSDaus_OVP
den_KFlux_CSVlat_LRA
cze_KDU−CSLaus_BZOfra_UMP
swi_EVP/PEVnet_PVV
slo_SMK−MKPpor_PP
it_FIlith_TS−LKD
bel_PPit_SVP
net_CUspa_PPpol_PR
bul_NFSBrom_PC
rom_UNPRslo_OLaNO
lat_NSLfin_KD
gre_LAOSlat_N_A
cro_HDZfin_PS
hun_Fideszit_UDC
ger_CSUit_NCD
gre_ANELsle_NSI
bul_VMRO−BNDlat_LKS
aus_FPOlith_TT
pol_PiSsle_SDS
cro_HDSSBpol_SP
den_DFrom_PP−DD
swe_SDcro_HSS
_Cslo_KDH
lith_LLRAswi_LdTlith_DK
ger_AfDpol_KNPtur_MHP
hun_JOBBIKfra_FN
cro_HSP−ASfra_MPFcro_HSPnet_SGP
swi_EDU/UDFswi_SVP/UDC
it_LNslo_SNS
bel_VBbul_ATAKA
tur_AKPgre_XA
it_Fdluk_UKIPlux_ADRger_NPD
−1.5 −1.0 −0.5 0.0 0.5 1.0
●●●●●
●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●
●●
Figure 4: Social left-right
12
Party Placements with 95% Credible Intervals
EU Position: Pro−EU parties
spa_ERClat_ZZSpol_PSLrom_PC
net_PvdAnet_CDA
lux_GRENGlux_DPire_Lablat_LRA
lux_LSAPsle_DeSUS
fra_PSrom_UNPR
bul_ABVden_V
mal_PLswi_GPS/PES
uk_PLAIDgre_DIMAR
fra_PRGfra_PRVswe_KDpor_PPsle_SD
cro_ORaHden_SDlat_N_A
rom_PSDit_VdA
por_PSit_NCD
est_EVEfin_SDPest_EERger_FDP
cze_CSSDbul_DPScro_HDZtur_HDP
aus_SPOslo_SMK−MKP
slo_Smer−SDuk_SNPlux_CSV
hun_MSZProm_PLR
cro_HSLSit_SVP
gre_Potamiire_FGfra_NC
it_CDfra_EELV
swe_Msle_PS
slo_SDKU−DSbel_SPAest_IRL
cro_SDPsle_SLSbel_PS
bel_ECOLOfra_AC
slo_MHfin_VIHRrom_PNL
aus_NEOSswi_SP/PSbel_Groen
spa_EApol_PO
spa_EAJ/PNVrom_UDMR
spa_CiUspa_CCit_UDC
sle_SMCaus_GRUNE
cze_KDU−CSLpor_PSD
net_GLcro_IDSbel_MRcze_SZ
uk_LIBDEMcze_TOP09
rom_PMPit_PD
ger_Grunenrom_PDLbul_DSBhun_E14
gre_PASOKsle_ZaABbul_SDSpol_SLDbel_VLDhun_DK
cro_HNSbel_FDF
aus_OVPswe_FP
_Bbel_cdH
bul_GERBest_SDEbul_DBG
gre_NDlith_TS−LKD
ger_SPDger_CDU
fin_RKP/SFPest_ERpol_RP
den_RVnet_D66
spa_UpyDspa_C's
lat_Vspa_PSOE
sle_SDSlith_LRLS
bel_CD&Vmal_PN
lith_LSDPspa_PPsle_NSIfin_KOK
it_SCcyp_DISY
−1.5 −1.0 −0.5 0.0 0.5 1.0
●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●
●
Figure 5: European Integration
13
Party Placements with 95% Credible Intervals
EU Position: Anti−EU parties
uk_UKIPgre_KKE
gre_XAden_FolkB
net_PVVswe_SDpol_KNPire_PBPA
it_LNfra_FNire_SP
fra_MPFbul_ATAKA
por_CDUhun_JOBBIK
swi_SVP/UDCit_M5S
ger_AfDcze_SVOBODNI
ger_NPDswi_LdT
fin_PSden_EL
cyp_AKELaus_FPO
swi_EDU/UDFden_DF
it_RCswe_V
uk_CONSfra_PG
lux_ADRtur_MHP
net_SPgre_ANEL
ire_SFslo_SNS
_Alux_DL
cze_USVITit_Fdl
net_SGPaus_BZO
sle_ZLbel_VB
aus_TeamStronachpor_BElat_LKSpol_SP
cze_KSCMbul_NFSB
hun_Fideszlith_TT
slo_SaScze_ODS
lith_DKbul_VMRO−BND
fra_Enscro_HSP
slo_OLaNObel_PP
ger_Linkeit_SEL
gre_SYRIZAswe_FI
tur_AKPpor_MPT
net_CUgre_LAOSbel_PVDA
den_LAit_FI
fra_MODEMpol_PiS
net_PvdDire_GPpol_PR
cyp_KOPswi_CVP/PVCcro_HSP−AS
swi_FDP/PLRfin_KD
spa_Amaiurswe_PIRAT
slo_NOVAswi_BDP
swi_EVP/PEVlith_LLRA
ger_DieTierspa_Podemos
cyp_DIKOspa_IU
spa_ICVswe_MPfin_VAS
uk_GREENlat_SDPS
lat_NSLcyp_EDEK
lith_LVZSbul_BBTden_SF
swi_CSP/PCScro_HDSSB
tur_CHPspa_BNG
ger_Piraten_C
est_EKfin_KESK
slo_Sietrom_PP−DDnet_50PLUS
cro_HSSslo_KDH
cze_ANO2011swi_GLP/PVL
ger_CSUlith_DP
uk_LABire_FF
net_VVDswe_SAPbul_BSP
cyp_EVROKOfra_UMP
hun_LMPbel_N−VA
swe_Cden_KF
cro_HL−SR
−1.5 −1.0 −0.5 0.0 0.5 1.0
●●●●●●●●●●●●●
●●●●●●●●●●●●●●
●●●●●●●●●
●●●●●●
●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●
●●●
●●●●●●●●●●●●●●●●●●●●●●
●●●●●●●●●●●●●●●●●
Figure 6: European Integration
14
An obvious question is whether or not the scale produced by the BAM procedure is in-
terestingly different from the unscaled placements in the CHES data. To answer this, we
sorted the data based on their rank orders on the two dimensions and plotted these against
each other. Thus, we created 2 new variables for each dimension that range from 1 to 262,
representing the lowest to highest values on each dimension. Next, we plot these two sets
of rank orders against each other. If there were no differences in the rank orders between
the BAM solution and the unscaled expert placements, the points would fall in a straight 45
degree line. To the extent that the rank orders differ, the points diverge from the ‘perfect’
fit. Figures 7, 8, and 9 display these comparisons.
UKIP
PBPAKKE
MPF
FN
XA
FolkB
KNP
SP
SD
PVV
LN
CDU
ATAKA
AfD
SVP/UDC
SVOBODNI
AKEL
EL PG
FPO
PS
DF
SFV
CONS
ANEL
RC
MHP
SP
Fdl
SGP
BZO
LKS
SP
TeamStronach
VB
BE
Fidesz
KSCM
NFSB
ZL DL
Ens
TT
HSP
SaSODS
DK
VMRO-BND
PP
Linke
OLaNO
SELSYRIZAFI
AKP
CU
LAOS
PVDA
LA
PiS
GP
PvdD
PR
HSP-ASKOP
CVP/PVC
Amaiur
FDP/PLR
PIRATEVP/PEV
NOVA
LLRA
KD
PodemosIU
DieTier
DIKO
ICV
MP
EDEK
VAS
LVZS
GREEN
HDSSB
BNGSF
EK
CHPPiraten
PP-DD
Siet
KESK
50PLUS
KDH
GLP/PVL
FF
ANO2011
DP
CSU
VVD
SAP
EVROKO
UMP
LMP
HL-SR
LABERC
N-VA
C
PSL
KF
PC
Lab
CDA
PS
UNPRDIMAR
PRG
GPS/PES
NA
PRV
KD V
PLAID
EER
PSD
EVE
DPS
FDP
HDP
PP
CSSD
SDP
SPO
MSZP
HDZ
SMK-MKP
Smer-SD
NCD
FG
PS
Potami
HSLS
PLRNC
SDKU-DS
EELV
M
SNP
SLS
SDP
MH
IRL
PSSPA
NEOS
AC
ECOLO
VIHR
PNL
SP/PSGroen
EAJ/PNV
GL
CiU
SMC
EA
POCC
UDMR
GRUNE
KDU-CSL
DSB
SZ
MR
IDS
SDSTOP09
GERB
Grunen
PASOK
DBG
PMP
DK
PDL
OVP
VLD
D66
PSD
HNS
ND
LIBDEM
SLD
FP
PD
SPD
cdH
FDF
CDU
TS-LKD
ERSDS
SDE
UpyD
RP
C's
LRLS
PSOE
CD&V
RV
LSDP
NSI
PP
JOBBIK
M5S
LdT
NPD
EDU/UDF
SNS
USVIT
ADR
MODEM
MPT FI
BDP
NSL
SDPS
BBT
CSP/PCSHSS
BSP
ZZS
PvdA
LRA
DeSUS
ABV
PL
GRENG
LSAP
DP
SD
ORaH
VdA
SD
NCD
PSCSV
SVP
CDE14UDC
ZaAB
V
RKP/SFP
PN
KOK
SC
DISY
0
50
100
150
200
250
Ran
k or
der
of E
con
Left-
Rig
ht: C
HE
S
0 50 100 150 200 250Rank of Econ Left-Right: BAM
Difference in rank < 20 Difference in rank > 20
Figure 7: Comparison of Economic left-right rank orders
15
UKIP
PBPA
KKE
MPFFN
XA
FolkB
KNP
SP
SD
PVV
LN
CDU
JOBBIKATAKA
AfD
SVP/UDC
M5S
SVOBODNI
NPD
EL
PG
EDU/UDFFPO
DF
SF
V
CONS
ANEL
RC
MHP
SP
SNS
USVIT
FdlSGP
BZO
LKS
SP
TeamStronach
VBFidesz
KSCM
NFSB
ZL
DL
Ens
TT
HSP
SaS
ODS
DK
VMRO-BND
PP
Linke
OLaNO
SEL
SYRIZA
FI
AKP
MODEM
CU
LAOS
MPT
PVDALA
PiS
GP
PvdD
PRFI
HSP-AS
CVP/PVC
Amaiur
FDP/PLR
PIRAT
EVP/PEV
BDPNOVA
LLRAKD
PodemosIU
DIKO
ICV
MP
EDEK
NSL
SDPS
VAS
LVZS
BBT
GREEN
CSP/PCS
HDSSB
BNG
SF
EK
CHP
PP-DD
SietKESK
HSS
50PLUS
KDH
GLP/PVL
ANO2011
DP
CSU
VVD
BSP
SAP
EVROKOUMP
LMP
HL-SR
LAB
ZZS
ERC
N-VA
C
PSL
PvdA
KF
PC
CDA
LRA
DeSUS
PS
ABV
UNPR
PLDIMAR
PRG
GPS/PES
GRENG
NA
PRV
DP
SD
KD
ORaH
V
PLAIDEER
PSD
EVE
VdA
DPS
HDP
SD
PP
CSSD
SDP
SPO MSZP
HDZ
SMK-MKP
Smer-SD
NCD
FG
Potami
HSLS
PLR
PS
NC
SDKU-DS
EELV
M SNP
SVPSLS
SDP
MH
IRL
PS
SPA
CD
NEOS
AC
ECOLO
VIHR
PNL
SP/PSGroen
EAJ/PNV
GL
CiU
SMCEA PO
CCUDMR
GRUNE
KDU-CSL
DSB
SZ
MR
IDS
SDS
TOP09
E14
GERB
Grunen
UDC
ZaAB
PASOKDBG
PMP
DK
PDL
OVP
VLD
D66
V
HNS
ND
LIBDEMSLD
FP
PD
SPD
cdH
RKP/SFP
FDF
CDU
TS-LKD
ER
SDS
SDE
UpyD
RP
C's
LRLS
PN
PSOE
CD&V
RV
LSDP
NSIPP
KOK
SC
DISY
LdT
AKEL
PS
ADR
BE
KOPDieTier
Piraten
FF
Lab
LSAP
FDPPS
CSVPSD
0
50
100
150
200
250R
ank
orde
r of
Soc
ial L
eft-
Rig
ht: C
HE
S
0 50 100 150 200 250Rank of Social Left-Right: BAM
Difference in rank < 20 Difference in rank > 20
Figure 8: Comparison of Social left-right rank orders
In these plots, we highlight (with red diamonds) parties that differ in rank ordering between
the two scales by at least 20 positions. These plots show that while there are some differences
in the rank ordering of the parties across the two scales, for a majority of the parties in the
data, these differences are not very large. Figure 9 shows that the EU dimension reflects
the most change in the rank ordering of the parties. We interpret this as meaning that
placements of parties on the pro/anti EU integration scale are the most vulnerable to DIF
and, as such, the vignettes have the biggest effect when estimating this scale, whereas the
social left-right dimension appears to be the least vulnerable to DIF.2
2The French party MODEM stands out as an extreme outlier on the EU dimension. This outlier is largelya function of several French experts placing two of the three EU vignette parties at the same position, thusmildly violating the vignette equivalance assumption. This also reflects that fact that several French expertsplaced MODEM, a pro-EU party, as being opposed to further integration
16
UKIPKKEMPFFNXAFolkBKNP
SDPVVLN
CDU
JOBBIKATAKAAfD
SVP/UDC
M5SSVOBODNILdTNPDEL
PGEDU/UDFFPOPSDF
SF
VANELRCMHPSPSNSUSVIT
ADR
FdlSGPBZOLKSSPTeamStronach
VB
BE
FideszKSCMNFSB
DL
EnsTT
HSPSaSODSDKVMRO-BNDLinkeOLaNOSELSYRIZAFIAKPCU
LAOSMPTPVDA
LAPiSGP
PvdDPR
FIHSP-ASCVP/PVC
Amaiur
FDP/PLRPIRATEVP/PEVBDP
NOVALLRAKDPodemosIU
DieTier
ICV
MPNSLSDPSVASLVZSBBT
GREEN
CSP/PCSHDSSBBNGSFEKCHP
PiratenPP-DD
SietHSS50PLUSKDHGLP/PVL
FF
ANO2011DPCSU
VVDBSPSAP
UMPLMP
HL-SRLABERCCPSLPvdAKF
PC
Lab
CDA
DeSUSPSABVUNPRPL
DIMARPRG
GPS/PESNA
PRVSDKD
ORaH
V
PLAIDEERPSD
EVE
DPS
HDP
SD
PPCSSDSPOMSZP
HDZSMK-MKPSmer-SD
Potami
HSLSPLR
PSNC
SDKU-DS
EELV
MSNP
SLS
MH
NEOSECOLO
PNLEAJ/PNVGL
CiU
SMCEA
PO
CC
GRUNEKDU-CSL
DSBSZ
MR
SDSE14ZaABPASOK
PMP
DKPDLOVPVLD
V
ND
LIBDEM
SLD
FP
PD
ERSDE
UpyDRPC's
PN
PSOE
RVPPSC
DISY
PBPASP
AKEL
CONS
ZL
PP
MODEM
KOP DIKO
EDEK
KESK
EVROKO
ZZSN-VALRA
GRENGLSAPDP
VdA
FDPSDPNCD
FG
PS
CSV
SVP
SLS
SDPIRL
PSSPA
CD
AC
VIHR
SP/PSGroen
UDMR
IDSTOP09GERB
GrunenUDC
DBGD66PSD
HNS
SPDcdHRKP/SFPFDFCDU
TS-LKD
SDS
LRLS
CD&VLSDP
NSI
KOK
0
50
100
150
200
250R
ank
orde
r of
EU
Pos
ition
: CH
ES
0 50 100 150 200 250Rank of EU Position: BAM
Difference in rank < 20 Difference in rank > 20
Figure 9: Comparison of EU positions
As mentioned above, the Bayesian framework produces measures of uncertainty around the
placements of the parties on the three dimensions. These are useful for determining whether
or not one party’s placement is meaningfully different from another. While most of the parties
that are placed in the middle ranges of these dimensions do not appear to be significantly
different from one another in terms of their positions, the parties near the endpoints of
the scales do seem to have meaningfully different placements. While we can informally
determine how different one party is from another by comparing the ranges of the 95%
intervals presented in Figures 1–6, the fact that we have posterior distributions for these
placements allow a more formal test of such potential differences. For example, the two
most extreme anti-EU parties according to the BAM solution are UKIP and KKE. According
17
to the unscaled CHES positions, these two parties are the seventh and fifth most anti-EU
parties in the data. This demonstrates that the British and Greek experts view the vignette
parties differently from one another, in that the rank ordering of these two parties actually
flips when the vignettes are considered. The question remains, however, as to whether or
not UKIP is meaningfully more anti-EU than is KKE. In order to answer this question,
we sampled 10, 000 draws from each parties’ posterior distribution and graphically present
these distributions in Figure 10. We then compare these two sets of draws and compute
the number of times that a draw from UKIP’s posterior distribution is more anti-EU than
the corresponding draw from KKE’s posterior. We find that 93% of the time (9, 300 out of
10, 000 draws), UKIP has a more extreme anti-EU position than does KKE. This is strong
evidence in support of the statement that UKIP is the most anti-EU party in our data, once
their placements are rescaled relative to the vignette parties and uncertainty is taken into
account.
−3.0 −2.5 −2.0 −1.5 −1.0 −0.5 0.0
0.0
0.5
1.0
1.5
2.0
Comparing EU Position Posteriors
Den
sity
UKIP
KKE
Figure 10: Comparison of EU positions
As an additional check on the validity of the BAM solutions, we examine the relationship
between the rescaled economic, social, and EU dimensions. We find that that dimensions are
related to one another in modest, yet expected, ways. The correlation between placements
18
on the economic and social left-right scales is 0.37, between the economic and EU dimensions
is 0.23 and between the social and EU dimensions is −0.33. Figures 11, 12, and 13 display
these relationships. The parties from Central and Eastern Europe are plotted in red to
highlight any differences in these relationships across East and West Europe. We also add
horizontal and vertical lines to separate the plots into four quandrants representing left and
right positions and pro and anti-EU positions. In Figure 11 we see that most parties fall
in either the economic and social left or the economic and social right quadrants, with no
obviously discernable differences between East and West Europe. As for the relationship
between economic and EU positions, we see the familiar inverted-U shape, discussed more
below. In Figure 13 we again see the inverted-U, with extreme parties on the left-right
dimension tending to be more anti-EU and those with moderate left-right positions tending
to be more pro-EU. Here, however, we see that there are very few parties in Central or
Eastern Europe that fall in the social left/anti-EU quadrant. In the CEES systems, anti-EU
parties tend to be more right-wing in terms of social policy. In the next section, we explore
these relationships in more depth.
19
SDS
BSP DPS
VMRO-BND
ATAKA
DSB
GERB
DBG
NFSB
BBT
ABV
CSSD
ODSKSCM
KDU-CSL
SZ
TOP09
ANO2011
USVIT
SVOBODNI
IRLEK
ER
SDE
EEREVE
MSZP
Fidesz
JOBBIK
LMPE14DK
LKS
ZZS
NA
SDPS
V
NSL
LRA
LSDP
TS-LKD
LVZS
LLRATT
DP
LRLS
DK
SLD
PO
PiS
PSL
RP
KNP
PR
SP
PSD
PC
PDLPNL
UDMR
UNPR
PP-DD
PMPPLRSDKU-DS
Smer-SDSMK-MKP
KDHSNS
SaS
MH
OLaNO
NOVASiet
SDS
SD
SLS
NSI
DeSUS
SMC
ZL
ZaABPS
HDZ
SDP
HSS
HSLS
HNSIDS
HDSSB
HSP
HL-SR
HSP-AS
ORaHPL
PN
DISY
EVROKO
AKEL
DIKO
EDEKKOP
PS
SPA
ECOLOGroen
MR
VLD
cdHCD&V
N-VA
FDF
VB
PVDA
PP
SD
RV
KF
SF
V
EL
DF
FolkB
LA
CDU
SPD
FDP
Grunen
Linke
CSU
NPD
AfD
Piraten
DieTier
PASOK
ND
SYRIZA
KKE
LAOSANEL
PotamiDIMAR
XA
PSOE
PP
IU
CiUEAJ/PNV
EA
ERCBNG
CC
ICV
UpyD
AmaiurPodemos
C'sPSPRG
EELV
UMP
FN MPF
MODEM
NCPRV
AC
PG
Ens
FF
FG
Lab
GP
SF
SPPBPA
RC
LN
UDC
FISVP
PD
SEL
CD
Fdl
M5S
SCVdA
NCD
CDA
PvdA
VVD
D66GL
SGP
SP
CU PVV
PvdD
50PLUS
CONS
LAB
LIBDEM
SNP
PLAID
GREEN
UKIP
CDU
PP
PS
PSD
BE
MPT
SPO
OVP
FPO
GRUNE
BZOTeamStronach
SDP
KOK
KESK
VAS
PS
RKP/SFP
VIHR
KD
V
SAPCFP
M
KD
MP
SD
PIRATFI
CSV
GRENG DPLSAP
ADR
DL
-1
-.5
0
.5
1
1.5
Soc
ial L
eft-
Rig
ht
-2 -1 0 1 2Economic Left-Right
East West
Figure 11: Social and Economic left-right BAM scales
SDS
BSP
DPS
VMRO-BND
ATAKA
DSBGERBDBG
NFSB
BBT
ABVCSSD
ODSKSCM
KDU-CSLSZ TOP09
ANO2011
USVIT
SVOBODNI
IRL
EK
ERSDE
EER EVEMSZP
Fidesz
JOBBIK
LMP
E14DK
LKS
ZZSNA
SDPS
V
NSL
LRA
LSDP TS-LKD
LVZS
LLRA
TT
DP
LRLS
DK
SLDPO
PiS
PSL
RP
KNP
PR
SP
PSDPC
PDLPNLUDMR
UNPR
PP-DD
PMP
PLR SDKU-DSSmer-SD SMK-MKP
KDH
SNS
SaS
MH
OLaNO
NOVA
Siet
SDS
SDSLS
NSI
DeSUS
SMC
ZL
ZaAB
PS HDZSDP
HSS
HSLS
HNSIDS
HDSSB
HSP
HL-SR
HSP-AS
ORaH PL
PN
DISY
EVROKO
AKEL
DIKOEDEK
KOP
PSSPAECOLOGroen MRVLDcdH CD&V
N-VA
FDF
VB
PVDAPP
SD
RV
KF
SF
V
EL DF
FolkB
LA
CDUSPD
FDP
Grunen
Linke
CSU
NPD AfD
Piraten
DieTier
PASOK ND
SYRIZA
KKE
LAOS
ANEL
PotamiDIMAR
XA
PSOE PP
IU
CiUEAJ/PNVEA
ERC
BNG
CC
ICV
UpyD
AmaiurPodemos
C's
PSPRG
EELV
UMP
FN MPF
MODEM
NCPRV
AC
PG
Ens
FF
FG
Lab
GP
SF
SPPBPA
RC
LN
UDC
FI
SVP
PD
SEL
CD
Fdl
M5S
SC
VdA NCD
CDAPvdAVVD
D66GL
SGPSP
CU
PVV
PvdD
50PLUS
CONS
LAB
LIBDEM
SNPPLAID
GREEN
UKIP
CDU
PPPS
PSD
BE
MPT
SPO
OVP
FPO
GRUNE
BZOTeamStronach
SDP
KOK
KESKVAS
PS
RKP/SFPVIHR
KD
V
SAP C
FP
MKD
MP
SD
PIRAT
FI
CSVGRENG DPLSAP
ADRDL
-2
-1
0
1
Ant
i/Pro
EU
Inte
grat
ion
-2 -1 0 1 2Economic Left-Right
East West
Figure 12: EU and Economic left-right BAM scales
20
SDS
BSP
DPS
VMRO-BND
ATAKA
DSB GERBDBG
NFSB
BBT
ABVCSSD
ODSKSCM
KDU-CSLSZ TOP09
ANO2011
USVIT
SVOBODNI
IRL
EK
ERSDE
EEREVE MSZP
Fidesz
JOBBIK
LMP
E14DK
LKS
ZZSNA
SDPS
V
NSL
LRA
LSDP TS-LKD
LVZS
LLRA
TT
DP
LRLS
DK
SLDPO
PiS
PSL
RP
KNP
PR
SP
PSDPC
PDLPNLUDMR
UNPR
PP-DD
PMP
PLR SDKU-DS Smer-SDSMK-MKP
KDH
SNS
SaS
MH
OLaNO
NOVA
Siet
SDS
SDSLS
NSI
DeSUS
SMC
ZL
ZaAB
PS HDZSDP
HSS
HSLS
HNSIDS
HDSSB
HSP
HL-SR
HSP-AS
ORaHPL
PN
DISY
EVROKO
AKEL
DIKOEDEK
KOP
PSSPAECOLOGroen MRVLD cdH CD&V
N-VA
FDF
VB
PVDAPP
SD
RV
KF
SF
V
EL DF
FolkB
LA
CDUSPD
FDP
Grunen
Linke
CSU
NPDAfD
Piraten
DieTier
PASOK ND
SYRIZA
KKE
LAOS
ANEL
PotamiDIMAR
XA
PSOE PP
IU
CiUEAJ/PNVEA
ERC
BNG
CC
ICV
UpyD
AmaiurPodemos
C's
PSPRG
EELV
UMP
FNMPF
MODEM
NCPRV
AC
PG
Ens
FF
FG
Lab
GP
SF
SPPBPA
RC
LN
UDC
FI
SVP
PD
SEL
CD
Fdl
M5S
SC
VdA NCD
CDAPvdAVVD
D66GL
SGPSP
CU
PVV
PvdD
50PLUS
CONS
LAB
LIBDEM
SNPPLAID
GREEN
UKIP
CDU
PPPS
PSD
BE
MPT
SPO
OVP
FPO
GRUNE
BZOTeamStronach
SDP
KOK
KESKVAS
PS
RKP/SFPVIHR
KD
V
SAPC
FP
MKD
MP
SD
PIRAT
FI
CSVGRENGDP LSAP
ADRDL
-2
-1
0
1
Ant
i/Pro
EU
Inte
grat
ion
-1 -.5 0 .5 1 1.5Social Left-Right
East West
Figure 13: EU and Social left-right BAM scales
21
3 Replication
A prominent paper using all three dimensions—economic left-right, social left-right (aka
GAL/TAN), and EU support—Marks et al. (2006) model support for European integration
in 2002 using the economic and social dimensions. In this section, we replicate the Marks
et al. (2006) analysis and extend it in two ways. First, we directly replicate the analysis for
2014. Second, we conduct the analysis with the rescaled versions of all three dimensions.
We find that the model holds up for 2014. In addition, the rescaled and cross-nationally
comparable measures provide more support for the reliability of these findings. For economic
left-right, the classic U-curve of support, where extreme parties on the economic left and
right are more likely to oppose the EU, still largely explains Euroskepticism in the most
recent expert survey. In contrast, for social left-right, the story is simpler, with socially
right-wing parties opposed while socially left-wing parties tend to favor the EU.
In this section, we use the rescaled dimensionality measures to evaluate the relationship
between economic left-right, social left-right, and support for European integration. We
start with the Marks et al. (2006) analysis, which emphasized the differences and similarities
between east and west on Euroskepticism. In both regions in 2002, extremism mattered,
such that parties in the middle on economic left-right supported the EU while the extremes
opposed; however, the major difference between the two regions is that in the west, right-wing
economic parties were also right-wing on the social dimension (correlation: 0.57) whereas
in the east, left-wing parties were more often on the social right-wing (correlation: −0.49).
While the correlations and associated patterns have weakened in the ensuing years, the
pattern still largely holds in 2014 with the rescaled measures, with a positive correlation
(0.40) between economic and social left-right in the west and a negative correlation (−0.20)
in the east.
22
In Figures 14a and 14b, we present the three dimensional space, with economic and social
left-right on the x and y-axes, while EU support is crudely measured as a trichotomous
variable.
-1-.
50
.51
1.5
Soc
ial L
eft-
Rig
ht (
BA
M)
-1.5 -1 -.5 0 .5 1Economic Left-Right (BAM)
Anti Neutral Pro
Note: n=141 parties. On the rescaled EU position measure ranging from -1.5 to 1.6, pro- parties havea score between 0.5 and 1.6, neutral parties have a score from -0.5 to 0.5, and anti- parties scorebetween -1.5 and -0.5.
(a) Western Europe
-1-.
50
.51
Soc
ial L
eft-
Rig
ht (
BA
M)
-2 -1 0 1 2Economic Left-Right (BAM)
Anti Neutral Pro
Note: n=102 parties. On the rescaled EU position measure ranging from -1.5 to 1.6, pro- parties havea score between 0.5 and 1.6, neutral parties have a score from -0.5 to 0.5, and anti- parties scorebetween -1.5 and -0.5.
(b) Central and Eastern Europe
Figure 14: Left-Right Dimensionality andPositions on European Integration
23
In the first two columns of Table 1, we first replicate the Marks et al. (2006) model for
the Western EU members. For 2002, Column 1 shows that economic right-wing parties
are more supportive of the EU than left-wing parties while social right-wing parties are
more Euroskeptical. Column 2 shows the effect of extremism. While the social right-wing
extremism variable just shows that the Euroskeptic trend is exaggerated on the social right-
wing extreme, the economic left-right extremism variable highlights the curvilinear nature,
the famous upside down U-curve. In short, the extremes on the economic left and right both
oppose the EU.
Columns 3 and 4 extend the 2002 model to the most recent CHES year. The same pattern
largely holds for 2014. Extremism matters for both economic and social left-right. As Marks
et al. (2006, 163) explain, “The reason for this is that the European Union is a centrist
project.” For our purposes, beyond the interesting consistency of the 2002 finding in 2014,
even after the many years of Euro crisis, the final two columns are the most significant.
Even with our cross-nationally comparable measures of these three dimensions, the same
substantive patterns and correlations are demonstrated in these data. This consistency
between the raw, unscaled, measures of dimensionality and our rescaled measures offers
more support for the extremism finding in the original paper.
Table 2 presents the same analysis conducted in the Central and Eastern European states.
Despite some differences among coefficients between east and west, the broad patterns (signs,
significance, etc.) are surprisingly consistent, suggesting that the geographic distinction
between the two regions is not as useful for this analysis as it once was.
Using the cross-nationally comparable dimensionality measures created in this paper, we
replicated and extended an earlier study of party-based Euroskepticism, illustrating that
ideological extremism is still a crucial factor to consider.
24
Tab
le1:
Anal
yzi
ng
Par
tyP
osit
ion
onE
U,
Wes
tern
Euro
pe
Rep
lica
tion
and
Exte
nsi
onof
Mar
ks
etal
.20
02
2002
Lin
ear
2002
Non
linea
r20
14L
inea
r20
14N
onlinea
rB
AM
Lin
ear
BA
MN
onlinea
rb
(S.E
.)b
(S.E
.)b
(S.E
.)b
(S.E
.)b
(S.E
.)b
(S.E
.)
Eco
nom
icL
eft-
Rig
ht
0.29
***
0.34
***
0.44
***
0.38
***
(0.0
2)(0.0
2)(0.0
2)(0.0
2)Soci
alL
eft-
Rig
ht
−0.
38**
*−
0.44
***
−0.
41**
*−
0.45
***
(0.0
2)(0.0
2)(0.0
2)(0.0
2)E
conom
icL
-RE
xtr
emis
m−
0.16
***
−0.
12**
*
(0.0
1)(0.0
1)Soci
alL
-RE
xtr
emis
m−
0.08
***
−0.
11**
*
(0.0
1)(0.0
1)E
conom
icL
-RB
AM
0.61
***
0.50
***
(0.0
3)(0.0
3)Soci
alL
-RB
AM
−0.
70**
*−
0.72
***
(0.0
4)(0.0
3)E
conom
icL
-RE
xtr
emis
mB
AM
−0.
53**
*
(0.0
4)Soci
alL
-RE
xtr
emis
mB
AM
−0.
85**
*
(0.0
5)C
onst
ant
6.06
***
7.00
***
4.86
***
6.53
***
0.23
***
0.68
***
(0.1
1)(0.1
0)(0.1
1)(0.1
1)(0.0
2)(0.0
3)R
20.
180.
450.
320.
530.
260.
44N
1310
1310
1428
1428
1418
1418
Note:
*p<
0.05
,**
p<
0.01
,**
*p<
0.00
1.O
rdin
ary
Lea
stsq
uare
sre
gre
ssio
ns.
Wei
ghte
dby
vote
.
25
Tab
le2:
Anal
yzi
ng
Par
tyP
osit
ion
onE
U,
Cen
tral
and
Eas
tern
Euro
pe
Rep
lica
tion
and
Exte
nsi
onof
Mar
ks
etal
.20
02
2002
Lin
ear
2002
Non
linea
r20
14L
inea
r20
14N
onlinea
rB
AM
Lin
ear
BA
MN
onlinea
rb
(S.E
.)b
(S.E
.)b
(S.E
.)b
(S.E
.)b
(S.E
.)b
(S.E
.)
Eco
nom
icL
eft-
Rig
ht
0.19
***
0.16
***
0.28
***
0.25
***
(0.0
2)(0.0
2)(0.0
2)(0.0
2)Soci
alL
eft-
Rig
ht
−0.
34**
*−
0.34
***
−0.
31**
*−
0.22
***
(0.0
2)(0.0
2)(0.0
1)(0.0
1)E
conom
icL
-RE
xtr
emis
m−
0.07
***
−0.
04**
*
(0.0
1)(0.0
1)Soci
alL
-RE
xtr
emis
m−
0.06
***
−0.
10**
*
(0.0
1)(0.0
1)E
conom
icL
-RB
AM
0.55
***
0.53
***
(0.0
3)(0.0
3)Soci
alL
-RB
AM
−0.
42**
*−
0.30
***
(0.0
3)(0.0
3)E
conom
icL
-RE
xtr
emis
mB
AM
−0.
11**
(0.0
4)Soci
alL
-RE
xtr
emis
mB
AM
−0.
54**
*
(0.0
6)C
onst
ant
6.65
***
7.33
***
5.81
***
6.16
***
0.45
***
0.60
***
(0.1
7)(0.1
5)(0.1
3)(0.1
1)(0.0
2)(0.0
2)R
20.
490.
600.
420.
540.
370.
42N
833
833
1197
1197
1197
1197
Note:
*p<
0.05
,**
p<
0.01
,**
*p<
0.00
1.O
rdin
ary
Lea
stsq
uare
sre
gre
ssio
ns.
Wei
ghte
dby
vote
.
26
4 Discussion
In this paper, we combined a series of anchoring vignettes that depicted hypothetical political
parties with Bayesian scaling techniques to produce cross-nationally comparable positions on
economic left-right, social left-right, and European integration for a wide range of political
parties in Europe. There was already growing theoretical and empirical evidence that the
economic left-right dimension travelled well across Europe (Bakker et al. 2014). The findings
we report in this paper indicate that social left-right and European integration also have a
high degree of pan-European comparability. This is important information, both because
these dimensions make up an increasingly important part of party competition in Europe,
and because it suggests that the items included in the Chapel Hill Expert Surveys measure
these dimensions quite effectively.
Our replication of a prominent study on the relationship between economic and social ide-
ology and European integration further illustrates this point. Although the research we
replicated used unscaled party positions from 2002, we find a similar structure to party
positions on European integration in the 2014 scaled data. Party positions on European
integration are structured and are systematically related to ideology on the economic and
social left-right dimensions.
Our findings also generate intriguing questions for future research. In particular, the fact that
meaningful differences between the scaled and unscaled positions on European integration,
although limited in number, tend to be concentrated at the extremes requires additional
examination. But in sum, the preliminary work that we present here supports the cross-
national comparability of key concepts for party competition in contemporary European
societies, and further bolsters the Chapel Hill Expert Survey as a valuable data source for
political parties in 21st century Europe.
27
References
Aldrich, John H. and Richard D. McKelvey. 1977. “A Method of Scaling with Applicationsto the 1968 and 1972 Presidential Elections.” American Political Science Review 71.1(March): 111–130.
Bakker, Ryan, Catherine De Vries, Erica Edwards, Liesbet Hooghe, Seth Jolly, Gary Marks,Jonathan Polk, Jan Rovny, Marco Steenbergen, and Milada Anna Vachudova. 2015. “Mea-suring party positions in Europe The Chapel Hill expert survey trend file, 1999–2010.”Party Politics 21 (1): 143–152.
Bakker, Ryan, Seth Jolly, Jonathan Polk, and Keith Poole. 2014. “The European CommonSpace: Extending the Use of Anchoring Vignettes.” The Journal of Politics 76 (04): 1089–1101.
Benoit, Kenneth and Michael Laver. 2006. Party Policy in Modern Democracies. London:Routledge.
Budge, Ian. 2000. “Expert Judgements of Party Policy Positions: Uses and Limitations inPolitical Research.” European Journal of Political Research 37.1: 103–113.
De Vries, Catherine E. and Sara B Hobolt. 2012. “When Dimensions Collide: The ElectoralSuccess of Issue Entrepreneurs.” European Union Politics 13.2: 246–268.
Hare, Christopher, David A Armstrong, Ryan Bakker, Royce Carroll, and Keith T Poole.2015. “Using Bayesian Aldrich-McKelvey Scaling to Study Citizens’ Ideological Preferencesand Perceptions.” American Journal of Political Science 59 (3): 759–774.
Hausermann, Silja and Hanspeter Kriesi. 2015. “What Do Voters Want? Dimensionsand Configurations in Individual-Level Preferences and Party Choice.” In The Politicsof Advanced Capitalism, ed. Pablo Beramendi, Silja Hausermann, Herbert Kitschelt, andHanspeter Kriesi. Cambridge Polity Press.
Hobolt, Sara and Catherine de Vries. 2015. “Issue Entrepreneurship and Multiparty Com-petition.” Comparative Political Studies 48.9: 1159–1185.
Hooghe, Liesbet and Gary Marks. 2009. “A Postfunctionalist Theory of European Integra-tion: From Permissive Consensus to Constraining Dissensus.” British Journal of PoliticalScience 39.1 (January): 1–23.
Hooghe, Liesbet, Ryan Bakker, Anna Brigevich, Catherine de Vries, Erica Edwards, GaryMarks, Jan Rovny, Marco Steenbergen, and Milada Vachudova. 2010. “Reliability andValidity of Measuring Party Positions: The Chapel Hill Expert Surveys of 2002 and 2006.”European Journal of Political Research 49.5: 687–703.
King, Gary, Christopher J. L. Murray, Joshua A. Salmon, and Ajay Tandon. 2004. “Enhanc-ing the Validity and Cross-Cultural Comparability of Measurement in Survey Research.”American Political Science Review 98.1 (February): 191–207.
28
King, Gary and Jonathan Wand. 2007. “Comparing Incomparable Survey Responses: Eval-uating and Selecting Anchoring Vignettes.” Political Analysis 15.1: 46–66.
Mair, Peter and Jacques Thomassen. 2010. “Political Representation and Government inthe European Union.” Journal of European Public Policy 17.1: 20–35.
Marks, Gary, Liesbet Hooghe, Moira Nelson, and Erica Edwards. 2006. “Party Competi-tion and European Integration in East and West. Different Structure, Same Causality.”Comparative Political Studies 39 (March): 155–75.
McDonald, Michael D., Silvia M. Mendes, and Myunghee Kim. 2007. “Cross-Temporal andCross-National Comparisons of Party Left-Right Positions.” Electoral Studies 26: 62–75.
Poole, Keith T. 1998. “Recovering a Basic Space From a Set of Issue Scales.” AmericanJournal of Political Science 42.3 (July): 954–993.
Ray, Leonard. 1999. “Measuring Party Orientation Towards European Integration: ResultsFrom an Expert Survey.” European Journal of Political Research 36.6: 283–306.
Rohrschneider, Robert and Stephen Whitefield. 2012. The Strain of Representation: HowParties Represent Diverse Voters in Western and Eastern Europe. Oxford: Oxford Uni-versity Press.
Rovny, Jan. 2014. “Communism, Federalism, and Ethnic Minorities: Explaining PartyCompetition Patterns in Eastern Europe.” World Politics 66 (04): 669–708.
29
Appendix. Vignette Text
Finally, we are going to present you with descriptions of 3 hypothetical parties and theirviews towards economic, libertarian/traditional, and EU issues. We would likeyou to placethese hypothetical parties on the following dimensions.
First, we would like you to place hypothetical Parties A, B, and C on the EU dimension.
Party A conceives the European Union as an intergovernmental organization inwhich member states, not the European Commission or the European Parliament,should be the dominant players. It rejects exiting the EU, but it wishes to reclaimstate sovereignty from the EU. On a 1-7 point scale with 1 being extreme anti-EUand 7 being extreme pro-EU where would you place this party?
Party B conceives the European Union as a supranational organization thatprovides Europeans with citizenship and a range of public goods. This party be-lieves the European Commission should become the government of the EuropeanUnion. On a 1-7 point scale with 1 being extreme anti-EU and 7 being extremepro-EU where would you place this party?
Party C believes that both member states and European institutions should playa vital role in EU policy making.The party is willing to pool national sovereigntyin the EU if this is efficient and feasible.European policy should be guided by sub-sidiarity, the principle that what can be better done at the national/subnationallevel should not be centralized. On a 1-7 point scale with 1 being extreme anti-EUand 7 being extreme pro-EU where would you place this party?
Next, we would like you to place the hypothetical parties on the economic dimension.
Party A advocates a social market economy with an emphasis on social justice,solidarity, and support for a welfare state. However, this party opposes stateownership, defends private property, and resists excessive intervention of the statein the economy. It believes there is a sharp trade-off between welfare spendingand economic competitiveness. On a 0-10 point scale with 0 being extreme leftand 10 being extreme right where would you place this party?
Party B views the equalization of life chances for all citizens as an importantgoal of government. It favors active government in regulating domestic and inter-national markets, and supports steeply progressive taxes to fund redistributivesocial programs. On a 0-10 point scale with 0 being extreme left and 10 beingextreme right where would you place this party?
Party C believes in small government. It favors minimal regulation of domesticand international markets, supports the privatization of many government oper-
30
ations, and opposes high taxes to fund redistributive social programs. On a 0-10point scale with 0 being extreme left and 10 being extreme right where would youplace this party?
Finally, we’d like you to place these 3 hypothetical parties on the libertarian/traditional
dimension.
Party A frames its policies around principles of social justice, grassroots democ-racy, and multiculturalism. The party favors same-sex marriage, active euthana-sia, and access to safe abortion. On a 0-10 point scale with 0 being extreme“Libertarian/postmaterialist” and 10 being extreme “Traditional/authoritarian”where would you place this party?
Party B favors non-discrimination legislation covering gender, race and sexualorientation, but opposes minority quotas. The party sees itself as a pragmaticparty that is willing to compromise if this is necessary to achieve its broad goals.On a 0-10 point scale with 0 being extreme “Libertarian/postmaterialist” and 10being extreme “Traditional/authoritarian” where would you place this party?
Party C emphasizes traditional family values, law and order, and the nation. Itopposes the legalization of same-sex marriage and the right to die. It believes thatthe government should be a firm moral authority on social and cultural issues.On a 0-10 point scale with 0 being extreme “Libertarian/postmaterialist” and 10being extreme “Traditional/authoritarian” where would you place this party?
31