new participation, new instruments: the construction of equivalent

1
NEW PARTICIPATION, NEW INSTRUMENTS: THE
CONSTRUCTION OF EQUIVALENT MEASURES OF
POLITICAL PARTICIPATION1
Gema M. García Albacete
University of Mannheim
[email protected]
Keywords: political participation, measurement, cross-cultural
equivalence, scaling, Mokken Scale Analysis
Gema García Albacete is a PhD candidate at the University of
Mannheim, Germany. She is enrolled in the Center for Doctoral
Studies in Social and Behavioral Sciences funded by the excellence
initiative of the German Science Foundation. Her dissertation
analyzes political participation of young people today in comparison
to former generations in Europe. Her research interests include
political behavior, public opinion and research methods in the social
sciences.
Abstract
Developments in the modes and levels in which European
citizens participate in politics urge for a revision of the
instruments used to measure political participation in cross
country comparative research. With this goal, the paper
explores the latent structure of political participation by using
the stochastic cumulative scale analysis procedure proposed
by Robert Mokken. Applying this procedure to the pooled
European Social Survey dataset shows the existence of two
dimensions and therefore the opportunity to construct two
scales. The first one includes actions related to the electoral
process; the second one contains protest forms of
participation. These two scales are tested for a large number
of countries in order to find out whether the scales can be
considered to establish an identity-set of actions with crossnational validity. To overcome the limitations implied in the
use of identical indicators cross nationally, in the next step the
identity-set is used as the basis to include national-specific
items in the instruments. As a result, the final measure of
1
Comunicación preparada para el IX Congreso de la AECPA. Grupo de trabajo 4.1, “La
participación política no electoral: de la acción individual a los movimientos sociales, retos
metodológicos y efectos políticos”. Málaga, 23 a 25 de Septiembre de 2009.
2
political participation developed here combines a crossnational, identical set of indicators for all countries as well as
a set of nation-specific indicators. In this way, non-identical
but equivalent scales are obtained for reliable and valid
measurement of political participation in cross-national
research. The external validity of the scales obtained is
explored by applying the measures to differences in political
participation across countries.
3
Introduction2
The development of citizens’ political participation in Western democracies
urge for a revision of the instruments used to measure political participation in
cross country comparative research. Social transformations in the forms and
extent in which citizens participate have lead to a misfit between the concept
of political participation and its measurement.
The following changes in citizen’s political involvement have been recently
documented: First, the traditional distinction between ‘conventional’ and
‘unconventional’ participation is obsolete (Teorell, Torcal, & Montero, 2007)
since involvement in what was considered ‘unconventional’ or ‘elite challenging’
political actions is now common place in European democracies (Topf, 1995a).
Second, emerging forms of political participation have been identified, such as
political consumerism and the use of new technologies with political purposes
(de Vreese, 2007; Delli Carpini, 2000; Micheletti, Follesdal, & Stolle, 2004).
Third, the agencies, or structures through which citizens are mobilized and
participate have also been transformed with the spread of new social
movements and advocacy networks (Norris, 2002). Finally, the weakness of
ties to political and civic organizations has resulted in an increase of
individualized patterns of participation to the detriment of collective action
(Pattie, Seyd, & Whiteley, 2004).
Despite these transformations and the rapid evolution and expansion of
political participation research since the 1950s, the instruments used in
political participation research have not been revised. Instead, researchers
tend to focus on specific forms of participation – electoral participation, protest
activities, political consumerism, etc. – usually measured by single items. This
strategy can result in misleading conclusions. As an example, we can revise
recent literature regarding young people political participation. While some
2
An earlier version of this paper was presented the workshop: Methodological Challenges in CrossNational Participation Research. The Hague, The Netherlands, 16-17 January 2009.
4
studies have called the attention to their low levels of political engagement
(Blais, Gidengil, & Nevitte, 2004; Pirie & Worcester, 1998; , 2000; Kay Lehman
Schlozman, Verba, Brady, & Erkulwater, 1999), others have concluded that
they are as involved as previous generations, but they just prefer to use new
forms of participation (Dalton, 2008; Gauthier, 2003; Micheletti, Follesdal, &
Stolle, 2004; O'Toole, Lister, Marsh, Jones, & McDonagh, 2003). Such
contradictory conclusions are due to the emphasis on single modes of
participation.
However, answering questions regarding citizens’ political involvement needs
instruments that integrate the social developments taking place. Or to put it in
a different way, it requires a higher coherence between the concept of political
participation – which implies a broader repertoire of actions citizens can get
involved in – and its measure.
Furthermore, comparative research poses another requirement. In order to
compare political participation across countries equivalent measures, which go
beyond the habitual identical measures, are needed. The degree to which
citizens actually participate and their modes of participation differ according to
country and context (inter alia, Anduiza, 1999; Ferrer, 2005; Kaase, 1989;
Teorell, Torcal, & Montero, 2007; Topf, 1995b). For instance, compared to
other European countries, participating in demonstrations is one of the most
common political activities in the South of Europe. Another example is the
higher degree in which citizens from Scandinavian countries get involved in
associations (Morales & Geurts, 2007, p. 138). The forms political participation
takes depend on the structure of opportunity and therefore, on the context. If
different – or non identical – modes of participation are available to citizens in
various countries, a comparable instrument to measure political participation
should take these differences into account.
Hence, the objective of this paper is to construct a measurement instrument
that includes the developments in political participation and has cross-country
validity. In order to do so, the latent structure of distinct forms of political
5
participation is explored in the following steps.
The first section briefly
describes some conceptual issues regarding the adequacy of considering
political participation as a latent construct. Second, using the stochastic
cumulative scale analysis proposed by Robert Mokken (1971) it presents the
existence of two scales in which citizens can be place according to their
participation across countries. The limitations involved in the use of identical
measurements are addressed by searching for equivalent measures following
the ‘identity-equivalence procedure’ introduced by Przeworski & Teune (1966).
Finally, the advantages of using equivalent measures for each country are
illustrated by their application to describe political participation across
countries.
An explorative approach to the dimensionality of political participation
Political participation research has been prolific during the last decades. Since the first
studies on voting behavior, the concept of political participation has evolved and
enlarged as the result of different waves of academic interest (Inglehart & Welzer,
2005; van Deth, 2001).3 The debate on the possible definitions of political
participation goes beyond the scope of this paper. From an explorative perspective,
what we need is a minimum definition, or better expressed “a demarcation line
between political participation and the rest of human behavior as a point of departure”
(van Deth, 1986, p. 262).
Political participation can be defined as “action by ordinary citizens directed toward
influencing some political outcomes” (Brady, 1999, p. 737). As Teorell et al. (2007, p.
336f.) argued recently, this definition – in comparison to more traditional
conceptualizations – allows including actions directly and indirectly aimed at
influencing the authoritative allocation of values in society.
We know the number of actions citizens have at their disposal to influence political
decision-making is numerous (van Deth, 2001) and this repertory has broaden in the
last decade (Norris, 2002). Therefore, in order to capture the whole concept of
3
See Morales (2004) or Schlozman (2002) for a review of the evolution of political participation research
from the study of electoral behavior to the multidimensionality of political participation.
6
political participation, an appropriate instrument to measure should be able to
integrate diverse political actions and modes of involvement.
The next question is how the different forms relate to each other. Political
participation is said to be a complex and multidimensional concept since individuals’
costs, motivations and consequences differ depending on the type of activity observed
(Verba & Nie, 1972). Since Barnes, Kaase et al. (1979) introduced the distinction
between conventional and unconventional forms of participation, it became a widely
used classification for subsequent studies given its ability to cluster the variety of
modes of participation and to articulate their different explanatory factors (Morales,
2004, p. 19). Therefore, the search for a measurement instrument of the concept
political participation must start with an exploration of its dimensionality and the
possibility of clustering the diverse activities in more than one dimension.
However, the observed differences regarding the factors that explain why some
people do – or do not – use specific modes of participation usually refer to their level
or difficulty, or in other words, to the degree of individual resources (motivation, time,
cognitive skills, money, etcetera) each of them requires. Already in the 1972, a latent
construct of political participation, or more exactly of “protest potential”, allowed
examining and locating individuals on an ordered continuum going from mild forms of
participation to more extreme actions (Marsh, 1974; , 1977). The basic question
underlying this dimension is “Think about protest. Generally speaking, how far are you
prepared to go?” (Marsh, 1977, p. 48). Building on the idea of studying political
participation as a latent continuum, an unidimensional ordered scale including both
conventional and unconventional modes of participation was successfully constructed
in the 1980s (van Deth, 1986).
The idea of ordering different modes of action contained in the concept political
participation is reinforced by the referred changes in citizens’ political participation. As
Norris has stated (2002), while political activists could be distinguished on whether
they would get involved in more institutional (or conventional) or non-institutional
modes of participation in the past, political participation has evolved and the domain
of political participation (or the repertoire of activities citizens have at their disposal)
has enlarged. Instead of only participating in protest or institutional forms, citizens
chose among the whole range of possible activities according to their resources and
interest in the topic, or coming back to Marsh’s concept, they will decide how far are
7
they prepared to go and act consequently.
Summing up, transformations in the way citizens participate and former research
suggest that an a priori distinction between forms of participation is not the adequate
strategy for constructing a measure of political participation. Instead, the strategy
followed here is to explore the latent structure of political participation. Allowing more
than one dimension, the possibility of ordering diverse actions that can be considered
as elements of the broad definition of political participation referred to before is
examined.
Strategy: examining the structure and dimensionality of the latent
concept political participation using Mokken Scale Analysis (MSA)
The explorative goal presented before can be articulated in the form of the
following question: can the varieties of actions included in our broad concept of
political participation be ordered in a meaningful way? The assumption that
different actions can be ordered according to their difficulty in a single (or
various) dimensions can be tested with the cumulative scale analysis
procedure proposed by Robert Mokken (1971). MSA is a combination of a
measurement model and a procedure that analyses each individual’s pattern of
responses to a set of items that are designed to be indicators of a single latent
variable.4 Contrary to other techniques such as principal components analysis
(PCA)5 or reliability analysis it includes an item parameter that shows how
items differ in their distribution.
The former two methods assume that the
items can be regarded as parallel, i.e. having the same frequency distribution
(the same mean and standard deviation). However, items differ in their
frequency distribution, which makes factor analysis of dichotomous data
difficult to interpret. Furthermore, the order of “difficulty” of the items often
has an important theoretical interpretation that should be taken in account.
It allows the researcher first, to investigate the dimensionality of a given data
and the construction of a scale (or a number of scales) from a given pool of
4
More on MSA can be found, inter alia, in Mokken (1971) Stokman & van Schuur (1980) van Schuur
(2003) and Wismeijer et al (2008).
5
See Teorell, Torcal & Montero (2007) for a recent example of an analysis of political participation
dimensionality using principal components analysis.
8
items. It is assumed that each subject has a certain, unknown value on the
latent dimension. For each item, the probability of a positive response
increases with that unknown variable. In this case, it assumes that each
respondent can be placed in a scale of political participation. The probability
that an individual participated in a specific action – i.e. attending a political
rally – will be higher if he or she participated in a less demanding – or easier –
activity such as voting. Second, MSA allows a confirmatory approach, that is,
the evaluation of a set of items as one cumulative scale. Therefore, we will be
able to examine the properties of such a scale. As a result, to test the
assumption that the diverse modes of participation can be ordered from easier
to more difficult actions in a scale of political participation. Finally, it offers the
possibility to test the scale (or scales) across different groups and thus, assess
whether the measurement instrument is valid cross-nationally.
A number of methodological arguments are behind the selection of MSA as an
instrument to explore and measure the latent concept “political participation”.
In comparison to other scaling methods (i.e. Guttman), it is probabilistic rather
than deterministic. While a deterministic scale will assume that all those
subjects who answered positively to a concrete item (i.e. contacted a
politician) also did it for less difficult items (i.e. voting), MSA takes
measurement error in account.
Another advantage is its bottom-up hierarchical clustering criterion. When
items do not constitute a homogeneous set of indictors of the same latent trait,
most scaling and other data reduction techniques – including reliability analysis
and factor analysis – use a top-down approach to find the best subset of
indicators, first investigating the whole set and then deleting the worst item (in
reliability analysis this would be the “alpha if item deleted” procedure). In
contrast, the search strategy in MSA is a bottom-up approach, in which the
homogeneity coefficient serves as the clustering criterion.
Furthermore, PCA always results in a number of principal components for any
set of items irrespective of whether these components are useful or not. On
9
the other hand, MSA is based on assumptions about trait dimensionality and
relationships between items and traits. These assumptions may be supported
by the data or not. The possibility of exploring the fit of the items to the latent
trait and to reject some of them is therefore another advantage.
Last but not least, contrary to factor analysis techniques – which assume the
confirmation of theoretical expectations – MSA can be used in an explorative
fashion. Given the changes in political participation repertories of western
citizens referred before, a detail exploration of the interrelations of different
modes of participation is more adequate.
The procedure comprises different
stages, which have been well documented before (Mokken, 1971; van Deth,
1986; van Schuur, 2003) Therefore, in the following section only some brief
comments regarding the procedure will be mentioned.
The expansion of high-quality data gathered in a large number of countries and
for different purposes allow for the referred revision of the instruments of
political participation. However, this expansion poses threats as well as
opportunities for research endeavors. On the one hand, as has been stated in
previous research, the growing number of actions considered as political
participation and the different batteries used in survey studies implies, in van
Deth’s words, the risk of moving from “the study of political participation
towards a theory of everything” (2001).
On the other hand, the rich data
available allows for a systematic analysis of the structure of political
participation in different contexts and across time. The data selected for the
application presented in this paper is the first wave of the European Social
Survey collected in 2002. Following the broad definition of political participation
presented before, the analysis includes a large number of items: voting,
contacting politicians, working for a political party, working for other political
organisation,
displaying
badges,
signing
petitions,
attending
lawful
demonstrations, donating money, participating in illegal protest activities,
media attention, discussing politics and membership in traditional political
10
associations (political parties and trade unions).6 Together with these political
actions, two other forms of participation are incorporated here. First,
involvement in ‘new’ political associations, these are: environmental, peace,
humanitarian and animal rights associations.7 Secondly, two indicators of what
have been labelled ‘emerging’ or ‘new’ forms of participation available in the
same survey battery are included: boycotting or buycotting products for ethical
and political reasons. Therefore, we can account for two of the changes in
participation referred before, the emergence of new forms of participation and
new channels of participation such as social movements beyond traditional
political institutions.
Each of the items was recoded into dichotomous variables with values 1
“participation” and 0 “no participation”. In the cases of no answers, these were
recoded into 0 in order to get a conservative measure including only those
respondents that actually declared to have participated in each activity.8
In comparison to other recent data sets, the ESS also includes a larger number
of European countries and therefore, allows examining the cross country
validity of the scales. In order to facilitate the interpretation of the results only
established democracies are selected. In total seventeen countries were
included.9
The first analysis conducted is the search for one or more scales in which the
diverse political actions can be ordered for all the countries pooled sample. The
interactive procedure works as follows. In this first step, based on item-pair
6
The exact wording used in the questionnaire is: “There are different ways of trying to improve things in
[country] or help prevent things from going wrong. During the last 12 months, have you done any of the
following? Contacted a politician, government or local government official; worked in a political party or
action group; worn or displayed a campaign badge/sticker, signed a petition; taken part in a lawful public
demonstration; boycotted certain products; deliberately bought certain products for political, ethical or
environmental reasons; donated money to a political organization or group; participated in illegal protest
activities”.
7
A single index was constructed including respondents who declared being involved (in any form) in any
of the four types of associations.
8
The only exception is those who didn’t have the legal age to participate in the general elections. These
cases were dropped out from the dataset (2628 cases in the pooled sample).
9
The countries included are: Austria, Belgium, Switzerland, Germany, Denmark, Spain, Finland, France,
United Kingdom, Greece, Ireland, Italy, Luxemburg, The Netherlands, Norway, Portugal and Sweden.
11
coefficients, Hjk, item coefficients, denoted as Hj, are defined that express the
degree to which an item is related to the other items in the scale. The total
score coefficient, H, express the degree to which the total score accurately
orders persons on the latent trait scale. It can be evaluated according to the
following rule of thumb: H < .3 “no scale”, .3 ≤ H < .4 “weak scale”, .4 ≤ H
< .5 “medium scale” and H ≥ .5 “strong scale” (Mokken, 1971). Negative
values are in conflict with the model (rejection of the item from the scale) and
usually interpreted as a sign of multidimensionality. That resulted to be the
case for items such as illegal protest and voting. Other items might also be
non-scalable (low Hj), for example, items such as membership in associations,
media attention and discussing politics dropped out in early stages of the
analysis. For the remaining items, a search for a next scale is started, and so
on.
The search for identical measures of the latent concept “political
participation”
Coming back to the first question posed, can the different modes of
participation be ordered in a theoretical interesting way? The result of applying
MSA to the pooled sample of European countries results in the existence of two
scales (see Table 1). The first one includes more institutional forms of
participation – voting, contacting politicians, donating money and working for a
political party – while the second one comprises protest activities – signing a
petition, displaying badges, attending lawful demonstrations and participating
in illegal protest activities –. Some items are rejected in the analysis due to
their low scalability (buycotting products, media consumption and discussing
politics) or monotonicity violations (boycotting products in the second scale).
The contents of the two final scales resemble to the classical distinction
between ‘conventional’ and ‘unconventional’ modes of participation (Barnes et
al., 1979) and corroborates the multidimensionality of the latent concept
political participation.
[Table 1]
12
According to the rule of thumb presented before, the resulting second scale
can be evaluated as a strong scale, while the first one is only a weak scale. In
both cases the reliability coefficients, measured by Rho are low (.55 and .57
respectively), which is due to the low number of actions included in each of
them and the lack of a uniform distribution in their difficulties.
A second step is the evaluation of the scales constructed from the pooled data
sample to each country separately; do these instruments allow the ordered
measurement of political participation across the seventeen countries selected?
The test of the scales for each country resulted in a number of violations of the
model which were solved by deleting from each scale the most problematic
item (donating money in the first scale and displaying badges in the second).
Each of the final scales contains three items and is a strong scale according to
its scalability coefficient H (see Table 2). Only in two countries remain items
that violate at least one of the assumptions of the model, these exceptions are
Denmark
in the
first scale and
the United
Kingdom
in
the second.
Nevertheless, as can be observed in tables 3 and 4, in all the remaining
sixteen countries the scales are strong. However, the reliability coefficients are
quite low in most of the cases, which is explained by the low number of items
in each scale and the lack of a uniform distribution of the items difficulties.
[Tables 2 and 3]
The search for equivalent measures of the latent concept “political
participation”
The results from MSA show that the latent concept of political participation can
be measured cross-nationally by means of two strong additive scales.
Therefore we can construct two scales composed of identical items that allow
the ordering of individuals and modes of participation across a large number of
countries.
However, the use of identical instruments imposes a number of weaknesses to
the valid measurement of concepts cross-culturally. First, it implies a high loss
13
of information, since available data regarding a large number of actions taken
by citizens to influence the allocation of values in society is left out. Secondly,
it entails the risk of not capturing the whole concept of political participation in
specific countries. In order to increase the reliability of our measures we can
go beyond the use of identical instruments by constructing equivalent
measures. With the same objective, the “identity-equivalence procedure”
proposed by Prezeworski and Teune (1966) was successfully applied by van
Deth (1986). By including country specific items in the already presented
identical-set of indicators, longer, more reliable, and more contextually
relevant instruments are acquired (Przeworski & Teune, 1966, p. 564).
Equivalent measures can be constructed in this case by building on the former
identical scales and testing the possible additions of the other items for each
country. Departing from the identity set of items from the previous analyses, a
test is conducted for each country separately. In this step the addition of each
of the items not included before is tested for each country. In addition to the
test of each single item, all possible combinations of items are tested. The
criteria for accepting or discarding an additional item are the following. First,
as in former steps, the final scalability coefficient (H) has to remain higher
than .30 and it has to pose no violation to the Monotonicity model. Second,
when several combinations comply with those requisites, the scale with higher
reliability is selected.
As can be seen in Table 4 the first scale was improved in thirteen out of the
seventeen countries by the inclusion of one item (in eight of them) or two
items (in five countries). Most of the resulting country specific scales are
medium scales, only in three of the countries the scales can be considered
weak.
Furthermore, in fourteen of the countries, the inclusion of country
specific items entails an increase of scales reliability.
[Table 4]
The second scale can be complemented with a larger number of items. In
14
seven of the countries, up to three extra modes of participation can be
included; two items in seven countries; one item in two countries; and only in
one of the countries the inclusion of any of the extra items implied a least one
violation
to
the
model.
The resulting
equivalent
measures
of
protest
participation are medium or strong scales, only one having an H value lower
than .40 (see Table 5). Scale reliability increases in all seventeen cases after
the inclusion of country specific items to the scales.
[Table 5]
Summing up, the application of the identity-equivalence procedure allows to
integrate more available information in the scales and to improve its properties
in terms of reliability. In comparison to the more limited identity sets, the new
equivalent scales capture better the concept of political participation in each
country.
Application
of
the
equivalent
measures
to
describe
political
participation across countries and groups of the population
Once obtained equivalent scales, they can be used in comparative research.
Scales can be computed by adding the number of actions (items) in which each
respondent has participated. Since the country specific scales differ in the
number of actions they contain, the sum is standardised by dividing it by the
number of items contained in each of the two scales for every country.
Therefore, the resulting indicators go from 0 (respondents that did not
participate in any of the actions) to 1 (those who participate in all the political
activities included in the scale for their country).
Figures 1 and 2 show the average involvement in each of the participatory
scales across countries. The first scale allows the comparison of the average
levels of involvement in institutional modes of participation (see Figure 1).
Three groups of countries result from this comparison. The first one includes
those with higher rates of participation: the four Scandinavian countries
15
included in the analysis and Belgium. The second group is formed by those
countries in which participation is lower, south European countries, France and
the United Kingdom. Finally, the remaining countries are located in a middle
position between the former two groups. The average position of each country
in the second scale, the one containing protest activities, shows a smoother
distribution but still significant differences among countries. Sweden and
Switzerland rank significantly higher than the rest of the countries. On the
other side of the spectrum, Greece and Portugal have significant low
participation in comparison to other countries (see Figure 2).
[Figures 1 and 2]
We
now
can
investigate
the
relationship
between
both
participatory
dimensions. Each country is situated in a two-dimensional space to examine
the relationship between the levels of involvement in institutional and protests
activities in Figure 3. It shows that high participation in one of the types do not
necessarily leads to low levels of participation in the other. Most countries
cluster in the upper left quadrant of the graph which indicates that political
participation in institutional modes of participation is higher than involvement
in the protest participation dimension. When looking at the situation of each
country in comparison to the others in a number of countries participation in
both institutional and protest modes is located in the medium of the space
(Austria, Luxemburg, Germany, Ireland and The Netherlands). In Sweden and
Norway, participation in both types of activities is very high in comparison to
other countries. In Belgium, Finland and Denmark, involvement in electoral or
institutional activities is comparatively higher. The same is true in Switzerland,
the United Kingdom and France with regard to protest participation. Finally, in
the south of Europe (Italy, Spain, Greece and Portugal) and particularly in
Greece and Portugal both participatory levels are quite low compared to other
European countries.
[Figure 3]
16
The structure of political participation, and the relationship between the two
scales, can also be examined by the correlation between both scales for each
country. Does the multidimensionality of the latent concept of political
participation imply a conflict between the two types of participation? Or in
other words, do citizens who participate in institutional activities refrain
themselves from participating in non-institutional or protest activities? Figure 4
shows the correlation between institutional and non-institutional participation
in seven European countries. Grey bars represent Pearson correlations using
the European Social Survey. In the seven countries, the relationship is positive
and statistically significant, ranging from the lowest in Finland (.17) to the
highest in Italy (.48). Therefore, we can confirm that both dimensions of the
latent concept of political participation are strongly related to each other 10. In
addition, we can examine the evolution of political participation structure. The
black bars in Figure 4 contain the replication of the former analysis with data
from the Political Action Study, collected in the late seventies.11 It confirms
again the positive relationship between both dimensions. Furthermore, the
comparison of both points in time shows the relationship is stronger nowadays
in almost all countries (with the exception of Finland). Therefore we find
evidence of the stated enlargement of citizens’ participatory repertories. The
probability that an individual participates both in Institutional and Noninstitutional political actions in higher now than in the seventies.
[Figure 4]
Finally, the scales constructed can be used to investigate the levels of
participation of different groups of the population. For instance, how do
participation in Institutional and Non-institutional activities change with age?
Or more specifically, does the relationship between
age and political
participation still conform to the curvilinear shape found in former studies? In
10
11
The same conclusion can be reached when analyzing seventeen countries.
It must be noted here that a former analysis of the structure of political participation using the Political
Action Study showed a one-dimension structure excluding voting was more adequate in the seventies
(van Deth, 1986) However, for illustrative purposes, the scales constructed here replicate the same
structure found in the present study with the European Social Survey.
17
order not to impose a functional form to the data, locally weighted polynomial
regression is used to explore the trend in political participation across age.
LOWESS smoothing scatter plots are presented in Figures 5 and 6, showing
that the relationship between involvement in political activities and age
confirms the expected pattern. Despite to a different extent in each country,
participation in institutional activities presents an inverted “U” shape. It
increases with age until reaching a peak between 40 and 50 years old, when it
gradually decreases. However, in Belgium and United Kingdom it remains
stable with aging (see Figure 5). On its part, participation in non-institutional
modes, in general, is higher among younger citizens and starts decreasing at
an earlier age (see Figure 6). It also presents higher variation in patterns,
which comes as no surprise if we take into account that protest activities are
more influenced by the context and therefore, period effects.
[Figure 5 and 6]
Conclusion
This paper departed from two challenges in comparative political participation
research. First, departing from the need to integrate societal developments –
such as the expansion of the participatory repertories citizens have available –
into measurement instruments, it has been shown how the enlarged number of
political actions can be ordered according to their difficulty in two different
scales with good scalability properties. The broad definition of political
participation used permitted an explorative approach to examine the structure
of the latent construct political participation and resulted successful in
identified two different dimensions, one containing activities related to
elections and involvement in traditional institutions and the second one
comprising protest actions.
The second challenge identified is the existence of country specific modes of
participation. In order to ensure the comparability of political participation
18
cross culturally the search for measurement instruments must go beyond
identical stimuli and come up with equivalent measures. Following the identityequivalence procedure introduced by Przeworski & Teune resulted in an
improvement of the scales in both theoretical and methodological aspects. It
allowed the construction of longer, and thus more reliable and contextually
relevant measures. The final measure of political participation developed here
combines a cross-national, identical set of indicators for all countries as well as
a set of nation-specific indicators.
Finally, the comparison of the average levels of participation for each country
clearly shows the applicability of the equivalent scales for cross national
comparative research. Countries can be situated in a two-dimensional space of
political participation to examine the extent and the form in which their citizens
get involved in politics in comparison to other European countries. In addition,
a preliminary analysis of the relation between both dimensions of political
participation has shown a strong correlation among them. Furthermore, some
other ways in which the scales can be used to investigate political participation
patterns and changes have been shown.
19
References
20
Table 1: Characteristics of the two scales constructed from the pooled data
set: items included, their frequency and scalability coefficients.
First scale
Second scale
H
.34
H
.44
Rho
.55
Rho
.57
Items contained
Mean Item H Items contained
Mean Item H
Voting
0.82
0.37 Signing a petition
0.28
0.54
Contacting politicians
0.18
0.31
Displaying badges
0.09
0.38
Donating money
0.09
0.31
Attending lawful demonstrations 0.08
0.37
Working for a political party 0.06
0.37
Illegal protest activities
0.44
0.01
Note. Data base used is the first round of the European Social Survey.
Table 2: Characteristics of the two cross-country identity scales constructed:
Items included, their frequency and scalability coefficients.
First scale
H
Rho
Items contained
Voting
.53
.43
Mean Item H
0.82
0.37
Second scale
H
Rho
Items contained
Signing a petition
Contacting politicians
0.18
0.31
Attending lawful demonstrations 0.08
0.37
Working for a political party 0.06
0.37
Illegal protest activities
0.44
Note. Data base used is the first round of the European Social Survey.
.57
.50
Mean Item H
0.28
0.54
0.01
21
Table 3: Testing identity set scales across countries.
Sample
Pooled sample
Austria
Belgium
Switzerland
Germany
Denmark
Spain
Finland
France
United Kingdom
Greece
Ireland
Italy
Luxemburg
Netherlands
Norway
Portugal
Sweden
.53
.64
.46
.56
.58
.48
.54
.49
.39
.45
.68
.52
.63
.54
.54
.44
.66
.50
.43
.59
.41
.50
.43
.33
.45
.37
.36
.37
.53
.42
.47
.45
.40
.40
.49
.40
First scale
Model violations H
-.57
-.51
-.51
-.65
-.56
Monotonicity .51
-.65
-.64
-.66
-.54
-.42
-.65
-.67
-.38
-.53
-.60
-.52
-.65
Second scale
Rho
Model violations
.55
-.48
-.47
-.49
-.51
-.46
-.68
-.44
-.62
-.42
Monotonicity
.57
-.53
-.67
-.47
-.39
-.47
-.58
-.46
--
Note. Data base used is the first round of the European Social Survey. Figures show the scalability (H)
and reliability (Rho) coefficients resulting of testing the two scales for each country sample.
22
Table 4: Items included and properties of the final equivalent first scales in each country.
Sample
Austria
Belgium
Switzerland
Germany
Denmark
Spain
Finland
France
UK
Greece
Ireland
Italy
Luxemburg
Netherlands
Norway
Portugal
Sweden
Ha
.64
.46
.56
.58
.48
.54
.49
.39
.45
.68
.52
.63
.54
.54
.44
.66
.50
Rhoa
.59
.41
.50
.43
.33
.45
.37
.36
.37
.53
.42
.47
.45
.40
.40
.49
.40
.53/.65
x
--
.39/.46
.43/.43
x
x
x
x
.45/.52
x
.41/.45
.40/.50
x
x
.53/.56
x
x
x
.43/.56
.40/.46
.38/.41
.50/.53
x
.40/.42
.40/.42
.64/.57
.39/.48
.58/.57
.38/.48
.40/.45
x
.53/.54
x
x
x
--
x
.39/.52
x
x
.36/.49
x
.47/.58
x
.47/.54
x
x
x
.48/.60
x
Hb
.53
.51
.43
.40
.39
.50
.49
.36
.40
.47
.39
.47
.40
.40
.44
.48
.50
Rhob
.65
.47
.56
.46
.52
.53
.37
.49
.42
.58
.48
.54
.50
.50
.40
.60
.40
Members
traditional
orgz.
Donating
money
Traditional
orgz. +
donating
money
Note. Ha and Rhoa represent, respectively, the scalability and reliability coefficients of the identical scale for that country. Hb and Rhob are the same coefficients
for the final equivalent scale. The remaining figures represent the scalability (H) / reliability (Rho) coefficients resulting of the addition of each item or
combination of items in the scale and for each country. The cells highlighted in grey indicate the item or combination of items added in the final equivalent
scales.
23
Table 5: Items included and properties of the final equivalent second scales in each country.
Sample
Austria
Belgium
Switzerland
Germany
Denmark
Spain
Finland
France
United
Kingdom
.54
Greece
Ireland
Italy
Luxemburg
Netherlands
Norway
Portugal
Sweden
H
.51
.51
.65
.56
.51
.65
.64
.66
.42
.65
.67
.38
.53
.60
.52
.65
Rho
.48
.47
.49
.51
.46
.68
.44
.62
.42
.57
.53
.67
.47
.39
.47
.58
.46
x
.39/.53
x
x
x
x
x
x
x
x
.40/.51
x
x
x
x
x
x
Badges
.45
.40/.51
.50/.55
.52/.57
.45/.51
.66/.74
x
.58/.67
x
x
.52/.60
.60/.70
x
.46/.48
.40/.48
.42/.63
.51/.50
Boycott
.38
.43/.55
.42/.52
x
x
.58/.70
.33/.48
.47/.62
.42/.49
.38/.59
.49/.57
.53/.66
.35/.55
.46/.51
.38/.46
.45/.60
.38/.47
x
x
.39/.52
x
.37/.46
.50/.67
x
.45/.60
x
x
.43/.56
.55/.66
x
.39/.52
x
.37/.61
.40/.47
x
.37/.56
--
x
x
x
x
x
x
x
.40/.58
x
x
x
x
x
x
x
.40/.60
--
x
x
x
x
x
x
x
.38/.58
x
x
x
x
x
x
x
.39/.63
--
x
x
x
x
x
x
x
.37/.61
x
x
x
x
x
x
.41/.60
x
.42/.57
x
x
.56/.74
x
.47/.66
.42/.55
.35/.61
.43/.61
.51/.72
x
.43/.55
x
.40/.66
.39/.54
x
.39/.59
x
x
.38/.50
x
x
.45/.65
x
.37/.66
.42/.61
.52/.72
x
.40/.56
x
x
.42/.52
x
.44/.65
.45/.61
x
x
.53/.74
x
.45/.71
x
.40/.69
.46/.66
.55/.66
.35/.60
.46/.62
x
.41/.65
.42/.56
x
.40/.67
.45/.64
.58/.60
x
.51/.74
x
.45/.73
x
x
.43/.68
x
x
.44/.62
x
x
.43/.60
x
.38/.62
--
x
x
x
x
x
x
x
.37/.63
x
x
x
x
x
x
x
.38/.64
--
x
x
x
x
x
x
x
.38/.65
x
x
x
x
x
x
x
x
--
x
x
x
x
x
x
x
.39/.67
x
x
x
x
x
x
H
.41
.40
.45
.58
.38
.56
.33
.45
.42
.40
.43
.52
.35
.44
.40
.40
.43
Rho
.60
.67
.64
.60
.50
.74
.48
.73
.55
.69
.68
.72
.60
.62
.66
.66
.60
Newtype
Buycott
Newtype +
badges
Newtype +
boycott
Newtype +
buycott
Badges +
boycott
Badges +
buycott
Boycott +
buycott
Badges +
boycott +
buycott
Newtype +
badges +
boycott
Newtype +
badges +
buycott
Newtype +
boycott +
buycott
Note. Ha and Rhoa represent, respectively, the scalability and reliability coefficients of the identical scale for that country. Hb and Rhob are the same coefficients
for the final equivalent scale. The remaining figures represent the scalability (H) / reliability (Rho) coefficients resulting of the addition of each item or
combination of items in the scale and for each country. The cells highlighted in grey indicate the item or combination of items added in the final equivalent
scales.
24
Figure 1: Average participation in Institutional activities in seventeen European
countries.
Figure 2: Average participation in Non-institutional activities in seventeen
European countries.
25
.4
Figure 3: Levels of institutionalized and protest participation in
seventeen European countries.
Scale 1: Institutional Participation
.1
.2
.3
Norway
Denmark
Finland
Belgium
Sweden
Austria
Luxembourg
Switzerland
Germany
Ireland
Netherlands
United Kingdom
Italy
Spain
Greece
France
0
Portugal
0
.1
.2
.3
Scale 2: Non-institutional Participation
.4
Figure 4: Pearson correlation between Institutional and Noninstitutional political participation in seven European countries
and two points in time.
.25
.27
Austria
.2
Britain
.4
.17
.17
Finland
.26
Germany
.34
.27
Italy
.49
.31
.35
The Netherlands
.2
Switzerland
.38
0
.1
.2
.3
.4
.5
.6
.7
.8
Political Action Study (1977)
European Social Survey (2000)
Note. All correlations are significant at the 0.01 level (2-tailed).
25
.9
1
26
Figure 5: Participation in Institutional activities across age in
seventeen European countries.
Lowess smoother
Belgium
Switzerland
Germany
Denmark
Spain
Finland
France
United Kingdom
Greece
Ireland
Italy
Luxembourg
Netherlands
Norway
Portugal
Sweden
0
.5
1
0
.5
1
0
Scale1: Institutional participation
.5
1
Austria
40
60
80
100
20
40
60
80
100
20
40
60
80
100
0
.5
1
20
20
40
60
80
100
20
40
60
80
100
Age
bandwidth = .3
Source: European Social Survey (2000-2002)
Figure 6: Participation in Non-institutional activities across age in
seventeen European countries.
Lowess smoother
Belgium
Switzerland
Germany
Denmark
Spain
Finland
France
United Kingdom
Greece
Ireland
Italy
Luxembourg
Netherlands
Norway
.5
0
1
.5
0
1
.5
0
20
60
80
100
Sweden
.5
1
Portugal
40
0
Scale 2: Non-institutional participation
1
Austria
20
40
60
80
100
20
40
60
80
100
Age
bandwidth = .3
Source: European Social Survey (2000-2002)
26
20
40
60
80
100
20
40
60
80
100