Idiosyncrasies of party identification in new

Idiosyncrasies of party identification in new democracies: the
case of Central & Eastern Europe
Radek Markowski
Polish Academy of Science
&
Warsaw School of Social Sciences and Humanities
I.
Introduction
In political reality and in the theory of political science the issue of the main mechanisms that
link citizens and their representatives has attracted a lot of attention. Numerous concepts and
phenomena had been covered to explain the genesis and the nature of this linkage. Among the
most profoundly studied is the concept of party identification. Notwithstanding its clearly
defined Michigan version, there exists a plethora of other ways of conceptualizing voter
attachment, loyalty, and identity with political entities called parties.
Apart from the importance of its individual level manifestations, the concept of party
identification has been utilized to depict and explain macro-system phenomena – the strength
of sociopolitical cleavages, the rootedness of the party systems and, more generally, their
institutionalization, as well as overall stability of polities, among other. In a nutshell, the
general lack of stable, socially rooted, long term attachments of citizens to their
representatives and parties is usually considered a problematic (to say the least) quality of a
democratic order.
II.
Theoretical remarks
Numerous publications devoted to the ontological essence of what party identification is and
what it ought to be have been published since the seminal invention had been presented by
the Michigan Four (Campbell et al 1960; Butler, Stokes 1969; Fiorina 1981; Schmitt &
Holmberg 1995; Achen 2002; Holmberg 2007)
----------In this paper an attempt is undertaken to describe and explain the phenomenon of party
identification, its determinants and correlates, in the new democracies of Central and Eastern
Europe. To understand this phenomenon in new democratic polities a reference object is
needed and the natural candidate is the Western European core stable democracies.
Moreover, in order to test some more detailed hypotheses we will utilize the Polish National
Election Study, as its "rich" (in term of variables) data allows to unveil relationships that
CSES data set – the set used in the analyses- falls short.
The main idea and goal of this paper is to compare the significance of political vs sociological
factors in creating the essence, strength and patterned relationships of the party identification
in new democracies. Numerous efforts have been accomplished so far, some of them
contributing substantially to our understanding of the ontology of the PID phenomenon
(Enyedi, Todosijevic, nd; Rose, Mishler 1998; Markowski, Grzelak 1999; Colton 2000; Tucker
2001; Tworzecki 2003; Klobucer, Miller, Gwyn 2002).
Several caveats are due though at this point: first, most of the analyzed data come from CSES
Module 2, which was in the field in the early and first half of 2000s, some countries feature
2001 year, other 2004, and for Poland we utilize the 2007 Polish NES. At any rate, we are
covering very young democracies, all of them about a dozen to fifteen years old. And only
some of them have a XX-th century legacy of being democratic; most don't have democratic
experience at all. It is thus very difficult to claim that what we depict and aim at explaining
here is the classical Michigan long-term attitude towards a political object. Second, taking into
account the very high volatility and governmental turnover in these countries we are on weak
grounds even to claim it is Fiorina's "running tally" (Fiorina 1981), a more rational evaluative
phenomenon springing from voters experience with party/incumbents accomplishments. It is
also debatable how close we are to Achen's (2002) idea of a current estimate of the benefit
between the parties. Third, we can however and we do unveil systematic responses to the
"classical" PID question, on whether there is a party that is close to the citizen and how close
this party is. Fourth, contrary to the spirit of the Michigan tradition, in the new democracies
we would expect – or at least hypothesize – a congruence between party identification and the
vote as well as between vote and positive affective attitude towards party-voted-for and,
finally, between pid and the affective component toward a preferred party. In a nutshell, we
are unveiling the (expected) patterned relationship within the party identity triangle: vote-pidparty affect. This is very different – I admit – from the "unmoved mover" idea, nevertheless in
the first ten or so year of democratic experience one has to rely on predictable, congruent
relationship between the three aspects of individuals' attitudes towards parties, if the ultimate
goal is creation of a rue PID. To put it differently, even if in the stable democracies the PID
concept should manifest its clear independence – over a longer periods of time and
incidentally - from the vote, in newly established democracies in the period of naissance, in the
creation period of attachment to a party some level of coherence (to put it plainly) between
the cognitive, affective and behavioral aspects of peoples' attitudes must be in place.
The main tasks of the empirical part of the paper is to arrive at the following conclusions:
1) What are the socio-demograhic, socio-communitarian and political factors that impact party
identification? What are the differences between Central & Eastern European (CEE) and
Western European (WE)? Can we unveil patterned differences that would explain those
differences and allow us to claim where in the "historical chain" do CEE countries and their
voters find themselves compared to stable democracies of the West?
2) What is the relative impact of the three mentioned above groups of factors in explaining
party identification, its strength and cohesion of party identity triangle relationship, the
affective, cognitive and behavioral ones?
3) Finally, utilizing the Polish NES, we will test some of the more detailed expectations
concerning the impact of individuals' social rootedness (measured by conformism, duration of
local domicile and organizational membership) and the impact of in-group individuals
positioning, their central vs peripheral within-group position
Classical sociological and part of the political variables need no explanation; some of the
political variables do call for discussion and all of the – broadly understood -"social
rootedness", its both aspect rootedness proper and individuals' within-group positioning.
Let us start with the latter, and with its theoretical justification. In social sciences we deal with
statics and dynamics and with factors that are pretty autonomous and other that are
ontologically interactive. We already know pretty much – back to our main topic – how
particular socio-demographic factors, such as education, age or income influence party
identification (……), even if the size of their impact varies temporally and spatially. At the
same time we still know much less (in the West) and almost nothing (in CEE) about the exact
mechanisms in the mezzo- and micro-social structures that affect political attitudes and PID.
In particular, the extent to which social embededness or rootedness of individuals matters for
formation of party identification and enhancing of party loyalties in time. Our hypothetical
expectation is that more socially rooted individual is more exposed to social pressures, which
in turn boosts – among other things – their party preferences and their identities with parties
in particular. In more operational terms we have to use whatever is (as of today) available in
the comparative studies; the ones we sue here is CSES and Polish NES . Unfortunately, CSES
lacks deliberately created variables that would tap the social rootedness of individuals I'd be
inclined to use in this paper. In such (pretty frequent) cases one has to use whatever is at hand
re-arranging or re-coding the available set of variables. As a consequence, we will look at the
extent to which particular citizen are located centrally in his/her group (defined by
educational, residential, age cohort or religiousness' criteria) in terms of Left-Right and/or of
other salient ideological dimension (for instance in Poland it is Solidarism—Liberalism).
Fortunately for Poland we can use other variables. Polish NES is composed – among other
enterprises - of a mass survey and a representative study of MPs elected at a given election1.
Its design allows for testing issue positioning of elites and masses on the same (approximately)
10 issues or policy areas. As a consequence, we are able to unveil the exact location of each
individual citizen vs his/her party (measured as mean position of its' MPs). In addition, the
Polish case study allows us to test the congruency of each individual vis a vis his/her
electorate, that is proximity of individual in relation to the mean of his/her fellow voters, the
ones supporting the same party. In other words we are able to measure one facet of political
representation considered to be the simplest way of judging the quality of representation;
Achen (1978) calls it "proximity" measure of the representative bond. To be sure, in this paper
I am not really interested in the quality of representation per se, be it of particular parties and
their electorates nor with the overall, systemic quality of representation of the median voter by
the government/incumbents. What we are interested here in is the "centrality" vs
"peripheriality" of individual's position compared to the mean of the respective groups s/he
certainly relates to, if not identifies with (which might be too demanding). This claim, of
course, might be considered debatable, nevertheless we assume that both MPs parliamentary
caucuses of parties as well as the electorate of a given party constitutes a – sort of – reference
group, or at least, an object of social identity. Moreover, although Polish NES allows to check
whether the individual correctly perceives his/her position within the electorate or in relation
to MPs, in this paper we will limit ourselves to measuring and analyzing the centrality of
individuals position within different groups, treating it as a crude measure of their clarity
(strength) of belonging.
1
The Polish NES survey is accomplished usually a week to two after the election day (our aim is to finalize the
fieldwork before the final composition of the government) on a representative for the adult population (meaning
eligible voters) of Poland, the sample size usually is 2000 respondents or more. The political elite study is
accomplished several months after the election and always covers MPs from all parties present in the Polish
parliament. The two – the mass survey and the elite study – have few modules that are comparable, for instance
the one that covers citizens and MPs attitudes towards
Finally, the analyses will account for three measures of – what I will tentatively call here –
social rootedness: (a) individuals' family residence duration in the locality one lives in (domicile
duration), (b) their organizational membership (trade unions and professional), and (c) one
indicator of social conformity2, which is expected to distinguish among individuals sensitivity
to social control and to the pressures of the majority.
An impact of five political independent variables is being tested in the analyses that follow:
(i) political efficacy, (ii) satisfaction with democracy, (iii) political knowledge, (iv) perceived
ideological 'stretch' of the party system, (v) parties' affective differential.3
Finally, let me stress that – for concerns of space – we do not test any economic-driven
phenomena and factors that obviously impact the party attachments of individual, in particular
if one tests the "running tally" approach to party identification.
III. Research Questions, Expectations, Hypotheses
In this paper I deliberately refrain from formulating precise hypotheses and submit also
research questions and 'expectations', due to the fact that the PID phenomenon is in statu
nascendi in new democracies and because of the partial novelty of the analyses.
The first hypothesis (H1) is that is that Western European countries will unveil more citizens
being (a) party identifiers and (b) more devoted, stronger identifiers than in the CEE
countries. Second, we expect (H2) that this is because of the sheer age of their party systems
and the number of electoral signaling game experienced by an average citizens in the two parts
of Europe. If this prediction is correct, we should then be able to manifest that individuals'
age makes a difference in determining party identification in the West, but not necessarily in
CEE, i.e. we predict that since voters – be it old or young - had more or less the same time in
new democracies to familiarize with parties, their accomplishments and their programmatic
promises, then age should not exert a differentiating effect among age cohorts, which should
2
3
The respective item wording reads: "Majority is always right, one ought to conform to it"
The first three are classical questions used in many surveys, and are repeatedly asked in CSES modules. Efficacy
is covered by two questions 1/whether "who governs makes a difference", and 2/ whether "whom we vote for
matters". Political knowledge is created as a simple index of three questions about political facts. Satisfaction with
democracy is the classical Eurobarometer question design. The remaining two – 'party affective differential' and
'ideological stretch' are interactively created. The first, 'party affective differential' is created as a difference
between the score of the most liked party and the mean affective score assigned by an individual to all remaining
parties in the system. The second, 'ideological stretch', is measured by the perceived position of the most leftlocated and the most right-located party in the system.
be seen in the case of the Western stable democracies. Instead, we hypothesize that the
citizens' of CEE party identification will be significantly more determined by short-term and
more ascribed status factors (religion, residence) than in the West (H3). The latter because of
high social mobility in the first decade of the transformation and because of newly acquired
thus unstable pattern of status inconsistency. As a consequence, we would expect Western
European party identities to be effectively affected by longer-term and more fundamental,
asset-like factors, like education for instance or by political factors like political efficacy and
political knowledge than the – more volatile - satisfaction with democracy (H4). In the same
vein, and more generally, we expect that Western European PID is more determined by
rational/cognitive factors than emotional/identitarian (H5), for instance more by education
and political knowledge than by religiousness and party affective differential and perceived
ideological polarization .
In the same spirit, we have similar expectations concerning the strength of PID among
Western and Eastern European citizen (H6)
No particular hypotheses are submitted as far as – what I tentatively call – "triple cohesion"
between vote, PID and affect towards the party is concerned. This is mainly due to
ontological and epistemological reasons. Let me put it this way: just as there exists a concept
of the 'normal vote', one would expect to have a similar construct , that is 'normal vote—pid
correlation'. Normal in the sense that it is meaningful, not very high bordering tautology and
at the same time not very low indicating no relationship between the two ontologically similar
concept. The problem is that we have no such a tested and re-tested indicator. Moreover, the
problem is related to the particular definition of PID we are inclined to analyze – if we embark
on the Michigan concept and moreover stay in US, we might allow for much weaker
relationship between the two. If we conceive it as a "running tally" and concentrate on
Europeans alone we should be more demanding concerning the strength of the association.
To conclude: keeping in mind the above problem, in what follows the empirical part will
present the extent to which in two parts of Europe there is cohesion between the three interrelated – cognitive, affective and behavioral aspects of party attachment.
Finally, we dwell into the issue of social rootedness of individuals. In this analysis one major
hypothesis will be tested, namely whether one can reasonably expect that PID is
predominantly determined by political preferences, attitudes and milieus rather than purely
social preferences and social embeddedness (H7). This general hypothesis is accompanied by
many detailed expectations: first, for instance, that this relationship will be more pronounced
in the more stable democracies of the West, and two, that it can be significantly modified by
the very individual level social constrains, in particular by the three facets of social rootedness
of individuals.
Some of the ideas are possible for testing only with the Polish case study.
The general prediction of H7 is tested by juxtaposing the party identification of those
individuals who are centrally located within their respective (different, social and political)
groups of "reference" with those occupying peripheral positions. The latter is in inverted
comas as we are not able to prove that the groups we treat as contextual locations of
individual are indeed "groups of reference" proper, as coined by social psychologist. We just
submit that the location within groups which is being tested in terms of individuals' centralperipheral position, are considered here important identitarian aggregates with which a
contemporary citizens must have an ability and capacity to relate to. Most of the groups
dwelled upon are social entities of unified structural position, clear aspirations and stable
values, such as age cohorts, university educated, residents of different size localities or
religious communities. The same, even more profoundly applies to political identities with
elites of parties people vote for. An obvious expectation is that those who are closer to the
mean position of a party on an issue or ideological stance is more likely to unveil party
identification than those whose distance is bigger; one may rephrase it in political
representation language – the better represented the individual by parties they vote for the
more likely they identify with them. In between the purely social locations and the purely
political (vis a vis parties) lie our estimates of locations within respective electorates. In this
instance we simply assume that there is – though vague and scattered – some knowledge
about politically like-minded people, and consequently we also expect that those who are more
proximate to the average fellow voter will more likely unveil party identification. This location
or – if you will – distance between individual's position and the mean of his/her electorate is
measured both on the ideological left-right scale as well as dozen issues (for Poland). To
conclude, this part of the analysis of the impact of centrality of individuals' location within
groups they belong to on their party identification, should allow us to say something about the
effect of social forces as well as social control and social ties on the development of PID as
compared to the sheer political representation impact measured by proximity indicator and to
a mixture of socio-political entities of electorates defined by issue/policy domains.
IV.
Empirical analyses
At the beginning let me emphasize that the analyses presented in this part are deliberately kept
at a simple level of aggregation and are consciously crafted to avoid complicated multivariate
designs.
The empirical part is divided into several parts. First, I present the basic distributions of party
identification in the CEE and WE countries. Second, the relationship between the three facets
of party attachments (PID, vote and affect) will be compared across the two parts of Europe.
Third, is devoted to simple multivariate analyses of determinants of party identification, its
strength, and cohesion between the three aspects of party attachments. In the fourth, details
of the impact of individuals' within group location, its centrality in various sociodemographic, public and political groups are presented. Fifth, analyses of the impact of social
rootedness of individuals is scrutinized. Overall discussion and conclusions follow.
The simple distributions of the positive answer to the classical question if there is a party
which one feels close to are presented below, in detail for the CEE countries in particular
years.
COUNTRY
CSES 1
Bulgaria
CSES 2
69
Czech Republic
81
72
Hungary
73
79
Lithuania
50
Poland
68
59
Romania
73
55
Slovenia
46
CEE
67
63
WE
75
73
The simplest distributions tell us that indeed the number of party identifiers is higher in
Western Europe than in Central and Eastern part. The difference though is not as high as one
might have expected, provided there are clear outliers in the East (Slovenia).
A more detailed picture is shown in Table 1, where the percentage of those who do not
identify with parties, as well as those who are definite about their party identification and
those who are less confident about it.
Table 1
about here
The bottom part indicates the average strength of party identifications in both parts of
Europe. Without discussing details we can claim H1 is confirmed: there are more party
identifiers in the West and the proportion of those who claim to be "very close" is also higher.
The overall distributions show quantitative differences, but not a qualitative one; at least these
simple aggregate percentages manifest that the differences between the new and stable
democracies are not as big as one might have expected. Moreover, if one accounts for the
systematically lower – on average by 10-15 percent – turnout in CEE countries than in the
most Western European ones, the difference in party identifiers is virtually non-existent.
It might however be that this is precisely why the turnout is lower.
Table 4.1 presents the results of logit analysis of determinants of having (as opposed to not
having) a party identification. The most important test concerning the socio-demographic
effects on PID refers to the role of age in new and stable democracies. As predicted by H2,
due to the fact that young and old citizens in new, fragile democracies have experienced more
or less the same "electoral democratic load", unlike in Western stable democracies where the
older cohorts have played the electoral game many times and have more experience with
political parties, age matters only in the West. In fact, in the overall model (after accounting
for political variables) manifest age exerts a direct impact on PID in CEE, but less significantly
than in Western Europe.
Table 4.1
about here
H3 turns out to be partly confirmed; indeed, PID in the new democracies of CEE is more
affected by income and simply place of residence and less than in the West by education.
Briefly, in stable democracies the more asset-like durable factor matter, whereas in CEE the
impact is exerted mainly by more ascribed factors and/or more incidental. Unlike our
prediction – religion does not influence party identification.
Political efficacy and political knowledge of individuals matters – as predicted by H4 - much
more in stable democracies, whereas in the CEE ones it is the perceived party affect
differential (big) and ideological polarization (small) that contributes to the occurrence of
PID. In other words, those CEEuropeans who although perceive the programmatic
ideological spectrum as pretty non-alternative and nevertheless unveil a clear affective
difference between a party of their preference and other parties in the system are the ones that
are more likely to be found among clear partisans (see as well Table 4.2).
Table 4.2
about here
The data presented in table 4.2 and the previous one, confirm the general expectations
submitted in H5 and H6, although it applies mostly to the effect of political knowledge,
education, but definitely less so to religiousness and residence.
The simple analyses of the relationships between pid, vote and party affect – for which we
didn't present any strict hypotheses - show that the differences for the two parts of Europe
are far from significant.
Table 1 and 2 about here
In all relevant cell of tables 1 and 2 one can see a moderate support for the claim there exist
more cohesion between the three facets of party attachment in Western stable democracies
than in the CEE new ones. At the same time we witness more active voters in the West, even
if their voting preference is at odds with their party identification or their affective party
preference. This result – had we seen more significant differences – would support the
hypothesis about more autonomous pid-vote relationship in stable democracies.
The impact of the social positioning of individuals within their respective groups on PID is
shown in the lowest two rows of tables 4.1 through 4.5 as well as in table 6.
Table 6
about here
The overall message derived from the analyses is hard to interpret as it both rejects and
supports hypothesis H7 at the same time. First, as expected the more classical political
representation in place the more likely party identifiers to follow. In case of five issues (as
different as crime, taxes, FDI), the relationship unveiled shows that the closer a person's issue
stances are to the mean of parliamentary caucus representatives' issue stances of the party s/he
votes for, the more likely s/he is to be a party identifier . The proximities on the remaining
four issues are statistically insignificant. Secondly, and this is the truly unexpected result, is
that concerning the impact of social group's location of individuals, based on the calculation
of ideological distances (for the whole Europe on the left-right and for Poland also for
'solidaristic –liberal' ideological space) between their self-placement and the mean of the sociodemographic groups they belong to. The groups were distinguished by age, education,
religiosity, and residence. The results showing that in all instances, across both regions of
Europe and separately for Poland, the more peripheral (i.e. distant from the group's mean) is
the position of an individual the more likely s/he is a party identifier. And the reverse, the
more proximate his/her ideological location vis-à-vis the mean for a particular group the more
like s/he is to be a non-identifier. Not only unexpected, but hard to interpret a result,
nevertheless almost universally true. At this point, without further in-depth analyses of the
phenomenon (which would clearly exceed the scope of this paper) one can only submit
speculation as to why this might be the case.
One might for instance speculate that the core (centrally located) representatives of the sociodemographic groups satisfy their identitarian needs or the in-group feeling by referring to the
purely social milieus. Most likely significant part of the population can satisfy their identity,
interests and representative needs by entirely relying on social networks, local organizations
and the like, without the need to enter the political domain. This is particularly true for those
who might consider themselves to be the core or true representatives of be it age or religious
group. If however a person belonging to a certain, pretty objectively defined group feels being
either by fate or by choice rejecting most or some of the principal group values and thus
consider him-/herself a non-conformist, s/he might have to adhere and utilize political
identities to satisfy his/her aspirations and interests.
One might be tempted to submit that PID plays a role of a functional substitute of – failed –
social relations and ties. This possibility certainly call for further in-depth analysis.
Finally, it shows that indeed what matters for party identification formation is the linkage in
the political domain, between groups of voters and their particular representatives, and much
less the positioning within social groups or even within issue publics defined by party
electorates.
In the last step, utilizing the Polish NES 2007, we try to control for the effects of three social
rootedness factors, described in the theoretical part – duration of family domicile,
conformism, and organizational membership. These three factors were aimed – among other
things - at helping us understand the real mechanism behind the unexpected effect of the role
of the marginality of the social group positioning on the development of PID, in this instance
in new democracies and unfortunately –due to data shortage – only in one case study.
Table 7
about here
Introducing the three control variables to the 20 equations that indicated significant
relationship with PID in table 6, now we were able to arrive at 18 "interpretable" models with
interaction terms. As the discussion of all of them would take a lot of space, in table 7
excerpts are presented – three models (#2, 7 and 13) that are indicative of similar patterns of
relationships between the set of independent variables and their interactions.
Before we discuss the models that manifest some or significant effects of the control variables
and their interactive effects, let us emphasize that for eight equations (out of 18 analyzed) all
three control variables – conformism, domicile durability and organizational membership – do
not matter at all. That is, irrespectively of the values of the three the general relationship
between in-group location and PID remains unchanged.
The pattern presented by model #2, which applies to three, socio-demographic groups'
location (defined by place of residence, education and age) allow us to conclude the following:
the overall pattern that the further away an individual is from the mean of the social group
one falls into the more likely the occurrence of PID holds, but this relationship is boosted if
the individual is a member of professional organization and the relationship is not influenced
by the remaining two control variables, i.e. domicile and conformism.
Pattern #7 pertains mainly to the relationship between individuals positioning vis a vis MPs,
party elite mean on two issues (FDI and KRUS – the farmers pension funds), and as we have
seen in table 6 the baseline models indicated that the closer the individual to his/her elite
mean the more likely occurrence of PID. However after introducing controls with
interactions, this pattern is modifies as follows: lack of organizational membership depresses
the baseline relationship with PID, while organizational membership enhances the probability
of PID. And again the relationship is independent of the values of the remaining two control
variables (conformism and domicile).
Pattern #13 applies mainly to individuals' position within two social groups aggregated by
age and education on the solidaristic-liberal ideological dimension and two other social
groups, defined by religiosity and place of residence on the issue of nomenklatura. In this
instance, the baseline model - which indicates that the further an individual is from the mean
of the so-defined social groups the more likely the PID to be found – is modified by the
duration of individual's family domicile: the longer the local social rootedness the weaker the
baseline relationship, i.e. the experience of strong social roots enhances the proximity effects
of within-group location, in other words stronger local roots of individual make them
probably more representative of particular social groups which also contributes to higher PID
what we would otherwise have expected.
Finally, most of the time conformism in our analyses depresses the probability of party
identification, and the reverse – the more nonconformism in place the more likely PID to
occur.
V.
Conclusions
1) The analyses presented in this paper are very preliminary, for this reasons they were
designed in a rather simple fashion with more emphasis on the details of the uncovered
relationships than aimed at offering universal generalizations.
2) One should emphasize that important socio-economic as well as detailed family life factors
were not accounted for. The socio-economic evaluations were set aside as they are
complicated and need a whole bunch of separate hypotheses and consequently analyses. Their
inclusion would have been (and will be) possible had the aim of the paper been to arrive at
conclusions concerning the mechanisms of party identification creation in new democracies,
which was not the aim of this paper. The detailed family life (not to mention long-term panel
studies) devoted to the topic are simply non-existing in the region.
3) Irrespectively of the limitations of the presented analyses some general and some detailed
results deserve mentioning, the most fundamental one being that 12 years or so, which is the
duration of democratic reality in the CEE countries, is definitely too short a period to test all
conceivable, well established theories of party identification. One needs at least one generation
to test some of the initial, very simple hypotheses in Michigan's tradition spirit concerning the
transgenerational transmission of political beliefs and values. Needless to say we need panel
designed studies that – so far – are missing from the region.
4) We were able to confirm some of the simple hypotheses submitted, most notably the ones
that indicate not so much the absolute difference between the stable and fragile democracies
differences (in fact in terms of simple distribution not that significant), but the underlying
patterned relationship between what we've designed to serve as a set of independent variables.
Most notably the way age of individuals impacts party identification and its strength. If we are
right, about this particular relationship in a decade from now we should be expecting a
decrease in the difference that age exerts on PID in the two parts of Europe.
5) Western European PID is also based on more stable asset-driven factors (education) and
seems to be embedded in a more rationally constituted political environment (the effects of
political knowledge and efficacy). Its Eastern counterpart seems to be more clearly linked to
less rational, more short term affective factors ( Significant effect exerted by "negative PID"
on the formation of the positive PID in Poland, an issue not discussed in the paper, but pretty
widespread across the CEE region – data available upon request).
6) The substantial part of the analyses has been devoted to the social rootedness of individuals
as factors contributing to the existence and creation of PID. The results obtained need further
in-depth analyses, nevertheless one can claim two phenomena to be valid: first, high quality
political representation measured as the proximity of individuals' position vis a vis the
parliamentary caucuses mean policy stances of the parties they vote for, confirms what we
know from stable democracies about its effects on PID. Second general result of this part of
the analyses indicating that relative marginal position within particular social groups enhances
PID, and this is a pattern true for both parts of the continent, even stronger in the West. This
relationship, at least in Poland is mitigated by the social rootedness of individuals, those who
are active or durable members (of be it the local community or professional associations) exert
a mitigating effect on unexpected relationship discovered here. On the other hand, so far a
speculative proposal (though supported by the clear way nonconformism enhances PID),
reads that individuals who might lack strong social identities and that are peripherally located
within the important socio-demographic groups they belong to, as an alternative value and
interest representative channel embark on party identification. The hypothesis that PID plays
a role of a functional substitute of unsuccessful social relations or ties has to be tested in the
future in more detail in deliberately designed research projects.
TABLES
Table 1. Percentage of party identifiers in CEE and WE and its strength
PID
0
1
2
no pid
weak pid - a little closer to a
party
strong pid - close to a party
DEGREE OF
CLOSENESS TO
THE PARTY
Total
1
2
3
7
8
very close
somewhat close
not very close
refused
don't know
Total
Central Eastern
Europe
%
36.0%
West Europe
Total
%
25.2%
%
28.4%
19.2%
29.2%
26.3%
44.8%
45.5%
45.3%
100.0% N =7624 100.0%
14.6%
56.8%
24.5%
.0%
4.1%
20.4%
52.9%
25.4%
.1%
1.3%
100.0% N=3865 100.0%
N=1988
N=2750
100.0%
1
5
19.0%
53.8%
25.1%
.1%
2.0%
N=1212
N=1598
100.0%
4
9
Table 2. Triple Cohesion: Basic links between the three aspects of party attachments –
pid, vote and party affect in stable and new European democracies (percentage
distributions)
Central Eastern
Europe
%
N
1
2
3
4
5
6
7
Triple (pid-vote-party affect)
cohesion
vote=pid, not triple
vote<>pid, irrespectively of affect
vote=like, no pid
vote<>like, no pid
pid=like, no vote
chaos - no vote, pid<>like or no
pid- no like
Total
Pearson Chi-Square Tests:
West Europe
Total
%
N
%
N
45.3%
2895
49.1%
9461
48.1%
12356
5.8%
5.4%
16.2%
4.1%
7.4%
370
348
1035
261
475
5.8%
8.6%
16.8%
5.7%
4.6%
1127
1655
3239
1106
889
5.8%
7.8%
16.6%
5.3%
5.3%
1497
2003
4274
1367
1364
15.7%
1003
9.4%
1808
10.9%
2811
19285 100.0%
25672
100.0%
Chi-square = 347.450
6387 100.0%
Sig. = .000 (df = 6)
vote_like3
Table 3. Double cohesions between the three facets of party attachments (each pair
separately)
1
2
3
4
like=vote
like<>vote
like. no vote
no like differ - vote or no
vote
Total
vote_pid3
Chi-square = 350.133
1
2
3
4
5
pid_like3
1
2
3
4
Total
%
69.7%
15.2%
11.1%
N
13447
2937
2139
%
68.2%
14.6%
13.4%
N
17519
3760
3430
3.1%
201
4.0%
762
3.8%
963
100.0%
6387
100.0%
19285
100.0%
25672
3265
348
863
1347
564
6387
54.9%
8.6%
7.5%
23.6%
5.4%
100.0%
10588
1655
1446
4546
1050
19285
54.0%
7.8%
9.0%
23.0%
6.3%
100.0%
13853
2003
2309
5893
1614
25672
3757
397
2032
201
6387
60.4%
7.4%
28.3%
4.0%
100.0%
11639
1423
5461
762
19285
60.0%
7.1%
29.2%
3.8%
100.0%
15396
1820
7493
963
25672
51.1%
5.4%
13.5%
21.1%
8.8%
100.0%
Sig. = .000 (df = 4)
like=pid
like<>pid
no pid. like
no like difer - pid or no pid
Total
Chi-square = 39.406
West Europe
Sig. = .000 (df = 3)
vote=pid
vote<>pid
no vote. no pid
vote. no pid
no vote. pid
Total
Chi-square = 366.392
Central Eastern
Europe
%
N
63.8%
4072
12.9%
823
20.2%
1291
58.8%
6.2%
31.8%
3.1%
100.0%
Sig. = .000 (df =3)
Table 4.1 Determinants of PID, socio-demographic, political and socio-interactive
(logit)
PID
Overal
(N= 7785)
Coe
Z
f.
(N= 17991)
Pseudo R2=.03
SOCDEM:
age
.01
edu-hi
.30
relig
-.15
residence
.02
income
.09
3.7
6
2.7
4
2.1
.47
2.3
7
efficacy
.10
polknowledg
e
.01
satdem
.41
affect_differe
nt
.26
ideolstretch
.00
2.8
.05
3.6
1
8.6
7
.15
3.9
.58
7.13
-.19
3.11
.06
1.79
.13
4.4
.01
.06
-.14
.07
.16
3.0
9
0.2
1
1.3
8
0.9
5
3.9
6
.14
.41
.35
.35
lr_dist_electo
rat
-.23
lr_dist_age
.30
1.62
.10
5.2
6
8.0
8
4.55
.07
2.27
.09
2.36
13.4
3
.18
.01
-.02
.24
.00
-.25
.40
.01
.43
-.11
.09
.21
(N=12053)
Pseudo R2=.03
1.8
.01
4.04
2.95
.36
4.21
1.37
-.15
3.67
10.4
2
-.01
1.29
0.18
.05
1.04
(N=5020)
Pseudo R2=.11
0.9
2
0.1
7
1.3
8
7.0
4
2.6
4
(N=20204)
Pseudo R2=.03
DISTANCE:
Pseudo R2
‘bold’ sig. p ≤ .05
.01
West Europe
Overal
Separate
(N=5131 )
models
Coe
Coe
Z
Z
f.
f.
(N=5938)
Pseudo R2=.03
(N= 17576)
Pseudo R2=.11
POLIT:
Cons.
Central Eastern Europe
Separate
Overal
Separate
models
(N=2654 )
models
Coe
Coe
Coe
Z
Z
Z
f.
f.
f.
.14
.27
.35
.34
-.21
9.8
.22
1.34
.08
3.3
5
7.1
9
.55
-.19
.02
.08
6.07
8.78
1.85
.55
2.53
(N=12556)
Pseudo R2=.11
2.66
.14
5.56
.15
4.82
.67
.13
0.74
.57
3.58
6.29
16.8
2
.44
3.22
.24
.29
8.4
.36
1.93
10.7
9
-.23
.01
0.45
.01
4.82
-.01
(N=4359)
Pseudo R2=.03
4.37
.01
(N=15845)
Pseudo R2=.03
-.25
6.13
-.25
.32
7.53
.38
2.08
.12
6.01
12.8
2
-.25
3.58
.42
8.12
Table 4. 2 Determinants of party identification strength (OLS)
Central Eastern Europe
West Europe
Europe
Overal
Separate
Overal
Separate
Overal
Separate
PID_DEGREE
(N= 6779)
models
(N= 2000)
models
(N= 4779)
models
Coe
Coe
Coe
Coe
Coe
Coe
t
t
t
t
t
t
f.
f.
f.
f.
f.
f.
(N= 14938)
(N= 5015)
(N= 9923)
SOCDEM:
R2=.04
R2=.04
R2=.03
5.6
3.4
6.1
age
0
3
3
.01
.01 4.45
.00
.00 1.71
.01
.01 5.59
2.7
6.3
edu-hi
7
5
.09
.23 8.96 -.01 .06
.18 2.94
.09
.21 8.27
1.9
religiosity
4 -.08 2.13 -.06 1.4 -.04 1.40 -.05 .98 -.07 1.18
-.06
2.0
residence
.02 .68
.04 1.76
.06
7
.06 5.34
.00 .14
.02
.66
1.2
2.7
12.0
income
5
7
5
.02
.06 3.61
.06
.10
.01 .25
.03 1.75
(N= 15809)
(N=4066 )
(N= 11743)
POLIT:
R2=.18
R2=.16
R2=.20
5.5
3.9
5.9
efficacy
.07
1
.09 6.47
.05
6
.08 7.47
.08
2
.09 6.40
polknowled
4.1
.03 .33
.17 1.66
.10 .79
.11 2.04
.13
.29 3.82
ge
3
4.5
5.5
satdem01
7
6
.19
.19 3.41
.06 .91
.13 3.96
.19
.12 2.49
affect_differ
7.1
12.5
8.4
17.2
6.4
10.8
.13
.17
.12
.15
.14
.19
en
5
3
7
6
9
0
ideolog
1.0
3.7
stretch
-.01
6 -.01 -.53 -.03
2 -.01 -.98
.00 .23
.00 -.07
(N= 17299)
(N= 3555)
(N= 13744)
DISTANCE:
R2=.06
R2=.05
R2=.06
lr_dist_elekt
5.9
3.6
4.1
or
-.10
4 -.13 6.24 -.10
2 -.13 4.98 -.10
6 -.12 4.73
8.0
11.1
6.0
7.8
10.1
lr_dist_age
0
1
7
1
9
.14
.22
.12
.17 7.41
.16
.24
Cons.
-.13
R2
.17
‘bold’ sig. p ≤ .05
-.03
.13
-.33
.21
Table 4. 3 Determinants of triple cohesions (pid=vote=party affect). Logit
Central Eastern Europe
West Europe
Overal
Separate
Overal
Separate
Overal
Separate
(N=7785)
models
(N=2654)
models
(N=5131)
models
Coef
Coef
Coef
Coef
Coef
Coef
Z
Z
Z
Z
Z
Z
.
.
.
.
.
.
(N= 16932)
Pseudo R2=.02
SOCDEM:
age
educ-hi
.01
.10
-.13
.02
.99
2.2
2
.48
.09
3.5
religiosity
residence
income
3.4
9
polknow
satdem01
aff_diff
idestretch
.09
.06
.44
.34
3.2
4
.32
3.8
6
9.5
2
-.01 -.22
DISTANC
E:
-.23
Cons.
Pseudo R2
-.14
.03
.10
.00
1.22
.01
6.09
.00
-.02
.21
2.28
1.29
-.19
.06
2.86
1.20
4.33
.10
2.46
.20
2.53
.11
5.1
8
3.6
2
-.17
.01
.16
.14
.28
4.98
1.77
.07
-.09
1.00
-.25
.19
-.01
.51
.30
.02
0.84
-.03
6.69
16.6
4
1.80
.36
.36
5.61
12.8
1
-.25
5.03
.29
5.62
3.5
4
1.9
1
2.0
1
.33
4.6
4
(N= 11501)
Pseudo R2=.02
.01
.17
-.10
.01
.07
3.6
6
1.5
5
1.1
0
.12
2.5
5
(N= 4822)
Pseudo R2=.12
.30
(N= 20207)
Pseudo R2=.02
lr_dist_el~t
lr_dist_age
.39
4.31
(N= 16800)
Pseudo R2=.12
POLIT:
efficacy
.01
(N= 5431)
Pseudo R2=.02
.34
3.8
8
-.04
6.7
2
11.
6
.00 -.12
-.30
-.22
.22
5.80
.23
1.48
.10
2.9
4
2.2
2
3.46
.39
5.71
-.10
.04
1.05
1.33
.08
3.28
(N= 11978)
Pseudo R2=.12
.36
3.9
6
.85
2.7
8
7.6
7
.01
.15
.10
.15
.54
(N= 4359)
Pseudo R2=.02
7.42
.01
.12
.51
4.72
3.29
.50
.37
3.76
11.1
3
.04
1.3
(N= 15848)
Pseudo R2=.02
-.20
.19
2.98
.12
3.0
2
2.4
5
-.27
4.46
.31
5.22
Table 4. 4 Determinants of pid-vote cohesion in stable and new democracies (logit)
Central Eastern
West Europe
Europe
Overal
Separate
Overal
Separate
Overal
Separate
models
(N=2654)
models
(N=5131)
models
(N=7785)
Coe
Coe
Coe
Coe
Coe
Coe
Z
Z
Z
Z
Z
Z
f.
f.
f.
f.
f.
f.
(N= 18011)
(N= 5985)
(N= 12055)
SOCDEM:
Pseudo R2=.02
Pseudo R2=.02
Pseudo R2=.02
2.1
age
.01 3.99 .01 5.30 .01
9 .01 3.45 .01 4.82 .01 5.22
.11 1.09 .43 7.14 .01 .05 .28 2.29 .15 1.54 .41 7.05
educ-hi
2.2
religious
3 -.16 1.85 -.14 1.22 -.11 1.12
-.16 2.17 -.15 2.39 -.18
1.1
residence
.01 0.33 .04 1.67 .08
3 .04 1.12 -.01 -.29 .03 1.24
3.9
income
4 .18 6.91 .03 .71 .08 2.68
.06 1.88 .11 4.21 .11
(N= 17599)
(N= 5043)
(N= 12566)
POLIT:
Pseudo R2=.10
Pseudo R2=.11
Pseudo R2=.11
1.1
efficacy
.11 4.01 .16 4.95 .07
9 .20 4.58 .12 7.22 .15 4.85
polknowledg
.01 0.05 .33 1.93 -.16 .39 .01 .02 .14 .56 .58 3.97
e
3.0
10.2
satdem
2 .35
1 .36 3.74 .32 5.67
.35 4.12 .42 5.09 .27
affect_differe
10.7
8.1
10.1
14.5
.25
nt
5 .31 14.3 .25
3 .32 9.51 .27
1 .33
3
Ideol_stretch
-.02 1.14 .00 0.17 -.01 .80 .01 .55 -.02 -.78 .01 .35
(N= 20207)
(N= 4359)
(N= 15848)
DISTANCE:
Pseudo R2=.03
Pseudo R2=.04
Pseudo R2=.03
lr_dist_elect
12.0
7.1
10.6
14.8
orat
-.31
2 -.32 7.43 -.31
8 -.33
5 -.32
3 -.33 6.17
10.9
12.1
6.1
19.5
10.0
lr_dist_age
.29
4 .39
1 .23
3 .33 9.34 .34
8 .40
3
Cons.
Pseudo R2
‘bold’ sig. p ≤ .05
1.66
.09
1.23
.09
1.97
.10
Table 4. 5 Determinans of pid-vote discrepancy in stable and new democracies (logit)
Central Eastern Europe
West Europe
Overal (N= Separate Overal (N= Separate
Overal
Separate
7785)
models
2654)
models
(N=1531)
models
Coef
Coef
Coef
Coef
Coef
Coef
Z
Z
Z
Z
Z
Z
.
.
.
.
.
.
(N= 18011)
(N= (5956)
(N= 12055)
SOCDEM:
Pseudo R2=.00
Pseudo R2=.00
Pseudo R2=.00
1.6
age
1.4
2.1
1
2
6
.00 -.83
.00
.00 .97
.01
.00
.00 .27
2.1
1.2
1.2
1.8
1.4
educ-hi
.26
1
.11
4
.08
6
.17
4
.22
9
.08 .61
1.2
religiosity
1.9
1.8
1.0
1.0
1
0 -.05
4
3
5
.08
.00 .12
.16
.06
.05
0.2
residence
.01 .27
.01 .52 -.07 -.91
.00 -.09
.01
8
.01 .62
1.3
1.3
0.3
income
7
8
6
.02 .66
.01 .56
.10
.06
.01
.00 -.14
(N= 17599)
(N= 5043)
(N= 12556)
POLIT:
Pseudo R2=.02
Pseudo R2=.02
Pseudo R2=.03
2.4
2.2
efficacy
7
3 -.01 -.63
-.06
.00 .09 -.03 -.76
.03 .63 -.04
polknowledge
4.4
0.3
1.0
2
-.02 -.07
.12 .78
.78
5
.01 .06 -.10
.18
0
satdem
3.2
1.8
0.1
3.7
-.03 -.42 -.34
1 -.35
1 -.20 -1.5 -.02
7 -.38
1
affec_differen
2.8
4.0
1.6
2.0
2.6
3.4
t
7 -.17
0 -.11
3 -.14
6 -.09
5 -.19
6
-.10
Ideol_stretch
5.3
4.0
.06 2.6
.00 -.14 -.01 -.25 -.04
5
.07
0
.01 .78
(N= (20207)
(N= 4359)
(N= 15848)
DISTANCE:
Pseudo R2=.01
Pseudo R2=.02
Pseudo R2=.01
lr_dist_elector
5.8
6.0
3.9
5.0
4.1
5.2
.27
.27
.34
.34
.25
.27
at
8
1
3
9
3
6
lr_dist_age
7.5
2.6
4.7
6.8
-.13 -3.6 -.15
2 -.11 -.91 -.18
5 -.13
1 -.14
4
Cons.
Pseudo R2
‘bold’ sig. p ≤ .05
2.17
.02
3.28
.05
2.07
.02
Table 5. Determinant of individuals’ vis a vis social groups’ placement on ideological
and issue stances on PID among Polish electorate. Entries: univariate logit regression
coefficients (Polish NES 2007)
Poland 2007
CEE
WE
Coef.
Z
Coef.
Z
Coef.
Z
Ideological distances from
socio-demographic groups:
LR_dist_relig_group
.30 6.50 ***
.25 4.31 ***
.32 7.78 ***
LR_dist_resid_group
.29 6.42 ***
.28 4.44 ***
.31 8.80 ***
LR_dist_educ_group
.28 6.34 ***
.28 4.35 ***
.31 8.84 ***
LR_dist_age_group
.27 6.08 ***
.27 4.58 ***
.32 .9.78 ***
SL_dist_resid_group
.20 5.62 ***
SL_dist_relig_group
.20 5.34 ***
SL_dist_educ_group
.19 5.23 ***
SL_dist_age_group
.18 4.65 ***
Distances on issues from
Respondents' party elite:
KRUS-issue_dist_elit
-.14 -3.92 ***
FDI_dist_elit
-.13 -3.53 ***
crime_dist_elit
-.12 -3.18 **
priv_dist_elit
-.12 -3.10 **
tax_dist_elit
-.08 -2.32 *
Other:
nomen_dist_resid_group
.07 2.26 *
crime_dist_elec
-.10 -2.23 *
nomen_dist_relig_group
.07 2.12 *
KRUS-issue_dist_age_group
-.06 -2.07 *
KRUS-issue_dist_elec
-.09 -2.03 *
FDI_dist_elec
-.09 -2.02 *
nomen_dist_age_group
.06 2.02 *
nomen_dist_educ_group
.06 1.95
Signif. codes:
.001*** .01**
.05*
Table 6. The impact of social group placements on PID, after accounting for control
social rootedness variables and interactions among Polish electorate 2007. (logit)
Model 2
Model 7
Model 13
coef
Z Sig. coef
Z Sig. coef
Z
sig
LR_dist_resid-gr-mean
.35
3.61
.000
---
---
---
---
SL_dist_educ_gr-mean
---
---
---
.23
2.93
KRUS-ISSUE_dist_elect-mean
---
---
---
---
---
----
.00
.01
.990
conformism
-.07
-.9
.366
-.12
1.51
.130
-.03
-.19
.849
orgmember
-.13
-.39
.700
.36
1.16
.240
2.09
3.02
.003
grand-roots
.35
1.51
.130
.29
1.33
.183
.30
.76
.450
LR_dist_resid_group*confomism
-.03
-.68
.500
---
---
---
---
---
---
LR_dist_resid_group*orgmember
.37
1.88
.060
---
---
---
---
---
---
LR_dist_resid_group*grand-roots
-.13
1.21
.227
---
---
---
---
---
---
.000 ---
-----
-----