REGIONAL ELECTIONS IN MULTI-LEVEL SYSTEMS
Charlie Jeffery
University of Birmingham, UK
Dan Hough
University of Nottingham, UK
This article analyses the dynamics of voting
behaviour and party competition in states where
elections to regional legislatures take place alongside
elections to national parliaments. In such ‘multilevel
systems’ voters cast votes for different bodies
that each have significant law-making powers.
Voters may well use different criteria in coming to
their voting judgements in the two different
electoral arenas, even though the parties that
compete for their votes may well be exactly the
same. Whether they do so, or whether instead they
follow the same cues in regional elections that
underlie their party choices in statewide elections is
a moot point and it is unclear how voters behave in
such multi-level contexts. What is, for example, the
relationship between voting behaviour in federal and
Land elections in Germany (or between statewide
elections and regional elections in Spain, or Austria
or the post-devolution UK)? Is there a regular and
predictable relationship between election results at
different levels? Can the results of regional elections
be explained by a simple transfer of the terms of
analysis used for statewide elections? Or how far do
territorial specificities pull voting behaviour at the
regional level away from the national ‘norm’?
Oddly, European political science has not
addressed these questions in any sort of depth. As a
result we lack analytical tools focused on
understanding the interplay of regional and
statewide elections. This article attempts to remedy
the deficit. It draws on literature developed to
analyse other aspects of voting behaviour to shed at
least some light on the interplay of regional and
statewide elections in multi-level systems, on how
far voters make different types of judgement for
different elections.
The article develops perspectives as it were ‘from
above’ and ‘from below’. In the first half of the
article we explore how far the statewide political
environment, focused in and on national
parliaments, shapes regional electoral outcomes. In
the second half we explore whether territory-
Abstract
★
European Urban and Regional Studies 10(3): 199–212 Copyright © 2003 SAGE Publications
0969-7764[200307]10:3; 199–212;035821 London, Thousand Oaks, CA and New Delhi, www.sagepublications.com
This article sets out to find ways of analysing the
relationship of regional and statewide electoral
processes in multi-level systems. First, we analyse a
number of ‘top down’ approaches with the aim of
assessing how and when statewide issues are
perceived as shaping regional election outcomes.
Second, we discuss a ‘bottom up’ approach in which
the importance of territorial politics can be
measured. Both of these approaches, although not
originally developed for use in this particular context,
provide at least initial techniques for mapping
out the dynamics of multi-level voting.They test for
the subordination of regional elections to the
electoral rhythms of statewide politics as well as
exploring how different patterns of voting behaviour
compare from region to region and from election to
election. Finally, we move on to apply these two
basic models to the cases of Germany, Canada and
Spain, illustrating that in contexts which lack deep
territorial cleavages, regional and statewide election
results are broadly similar. However, in territorially
heterogeneous environments, this pattern of
subordination of regional elections is broken up by
territorially specific influences.
KEY WORDS ★ multi-level voting ★ territorial
politics ★ voting behaviour
specific issues at the regional level imbue regional
elections with dynamics that are uncoupled from the
statewide arena. In each case we first review how the
existing literature can help us address these
questions before then presenting empirical data
from Germany and Spain (and in the second half
also Canada) about the relationship between
statewide and regional election results. The aim is to
enable comparative, cross-country and interregional
analysis of multi-level voting.
For reasons of simplicity we take some short
cuts. We work throughout with aggregate-level data
and do not seek to contextualize the election results
that we analyse. We do not, therefore, take into
account the ‘hard’ structural variables (see Johnston,
1980: 131–2) which help shape election results such
as electoral systems or the constitutional status of
particular regions. We also do not take into account
other contingent factors – issues, personalities – that
naturally affect the outcomes of all elections. What
we do attempt to do is locate some starting points
for the comparative analysis of electoral dynamics in
multi-level systems.
The view ‘from above’: regional elections
as second-order elections
An obvious starting point in exploring the view
‘from above’ – whether statewide issues shape
regional election outcomes – is to look at the
literature on ‘second-order’ elections. This was a
terminology coined after the first direct elections to
the European Parliament (EP) in 1979. The 1979 EP
results revealed a pattern in which, broadly
speaking, parties currently in national government
underperformed ‘in Europe’. Writing in the
immediate aftermath, Karlheinz Reif and Hermann
Schmitt (1980) proposed that the 1979 EP elections
should be viewed not as European elections, but
rather as ‘second-order’ national elections. In other
words, they presented voters with little extra
stimulus than that with which they were already
confronted in the ‘first order’ of national elections,
those through which national governments are
elected and seats in the national legislature are at
stake. For Reif and Schmitt all other elections,
including ‘by-elections, municipal elections, various
sorts of regional elections, those to a second
chamber and the like’ (Reif and Schmitt, 1980: 8),
have ‘less at stake’ and have to be regarded as
‘second-order’.
The basic premise of Reif and Schmitt is
therefore that outcomes of second-order elections
are determined by the political situation in the firstorder,
i.e. statewide, political arena (Van der Eijk et
al., 1996). Reif made the point more forcefully after
the next set of EP elections in 1984: ‘What is
important is the political situation of the first-order
arena at the moment when the second-order election
is being held’ (Reif, 1985: 8).
Reif and Schmitt (1980: 9–10) set out the
following key propositions about second-order
elections:
• With less at stake than in first-order elections
turnout will be lower
• Where less is at stake, voters may ‘experiment’
with smaller or new parties close to their
preferences, but for which votes would be
‘wasted’ in the first-order arena
• Crucially, the government parties in the firstorder
arena are likely to lose support at midterm,
and opposition parties to gain support
• But while voters in these ways may ‘punish’
incumbent parties of national government, they
do so in the knowledge that they can return to it
‘when it really matters’, i.e. at the next first-order
election.
The notion of first/second-order elections is now
firmly established in the analysis of European
election results. What is a little surprising is how
few applications of the first/second-order
framework have been made to the other kinds of
second-order election Reif and Schmitt set out. One
exception is a series of articles comparing British
local government and EP elections during the 1990s.
These both confirm the general thrust of Reif and
Schmitt’s propositions, but also differentiate the
‘rank order’ of local and EP elections. ‘Some second
order elections are more second order than others’,
as McLean et al. (1996: 4) put it in comparing 1994
local and EP elections. Heath et al. (1999: 391) were
more precise five years later: ‘If the elections to the
European Parliament are to be regarded as secondorder,
then we might think of elections to local
councils as “one and three quarters order”’. These
differences had to do with the amount at stake,
European Urban and Regional Studies 2003 10(3)
200 EUROPEAN URBAN AND REGIONAL STUDIES 10(3)
which was felt to be higher (or at least more directly
appreciable) in local than EP elections, and was
reflected in higher turnout, a higher degree of
concern about who won, and a lesser likelihood to
vote in accordance with national/first-order issues
(McLean et al., 1996: 18; Heath et al., 1999: 406).
There is an immediate ‘read-across’ from these
analyses of British local elections into the
relationship between regional and statewide
elections. Just how ‘second-order’ are regional
elections? What is the relationship between the
regional and statewide electoral arenas in terms of
turnout, small parties and national government
popularity? Is there a regular pattern in which
regional elections routinely act as vehicles for the
‘punishment’ of national government parties?
Regional elections in the German mid-term
A little-rehearsed method for investigating such
questions was developed in 1977 by the German
political scientist Rainer Dinkel (1977). Dinkel
aimed to test the hypothesis that the parties in
government at the federal level inevitably suffer
drops in popular support in German regional
(Land) elections. He developed a simple test based
on the notion of ‘expected vote share’ at Land
elections. This was a simple average of party vote
shares in that Land at two successive federal
elections. If the federal government parties scored,
say, 50 percent in a Land in one federal election and
46 percent in the next, then their ‘expected vote’ for
any Land election in the intervening period in that
Land would be 48 percent. Dinkel found as an
almost invariable rule that the federal government
parties did less well than ‘expected’ in any Land
election than they had done in that Land in the
preceding or succeeding federal elections. The
regularity of this phenomenon was understood as
the ‘punishment’ of federal-level incumbency.
‘Land elections have long since become arenas
where citizens give their verdict on the performance
of the coalition parties at the federal level.’ Land
elections were (in anticipation of the later
terminology of Reif and Schmitt) ‘subordinate
elections […] systematically influenced by the
superordinate constellation in the Bundestag’
(Dinkel, 1977: 348).
Drawing on work on the ‘electoral cycle’ of
presidential vis-a-vis congressional elections in the
US, Dinkel also tested for cyclical variations in the
underperfomance relative to ‘expected’ vote share
over time between two federal elections (cf. Jeffery
and Hough, 2001; Hough and Jeffery, 2003a; 2003b:
82). The further away that a Land election took
place from a federal election, the worse the federal
government parties performed: the best chances for
federal government parties to maximize their vote
share was if the election took place either at the
beginning or the end of the federal electoral cycle
(Dinkel, 1977: 351). In other words, federal
government parties did not just underperform in
Land elections, they underperformed most at mid-
term.
Dinkel’s was not a complex model. It
(consciously1) had just one main explanatory
variable – distance in time from the nearest federal
election – and focused just on the performance of
those parties which happened to be in federal
government during a particular legislative period (in
other words, treating coalitions led by different
parties in different periods in an undifferentiated
way). It also lacks predictive capability. The result of
the federal election after any round of Land
elections is needed in order to calculate the trend in
support for the parties. While this is logical in terms
of seeking cyclical effects, Dinkel can as a result
offer only a retrospective evaluation of party
performance across tiers of government: Land
election results since the last federal election cannot
be evaluated for their cyclical characteristics until
the next federal election has taken place. Despite
these qualifications his model has a genuine
heuristic value. It is to our knowledge the only
attempt systematically to model the relationship
between national and regional elections. And the
fairly unambiguous empirical data which generated
it match closely the expectations of related work in
the first/second-order literature on ‘punishment
effects’ and the subordinate status of regional
elections to the first-order electoral arena.2 We
therefore use it in the next section to revisit the data
on Germany, but also to compare it with equivalent
data from Spain.
European Urban and Regional Studies 2003 10(3)
JEFFERY & HOUGH: REGIONAL ELECTIONS 201
Comparing Germany and Spain
Applying Dinkel’s method to Land election results
in the Federal Republic of Germany from 1949 to
1990 reveals a mild but distinctive cyclical pattern
(see Figure 1). The governing parties at the federal
level perform better in regional elections at the
beginning and at the end of the national legislative
cycle, whereas in ‘mid-term’ they perform less well.
Only very rarely do they manage to poll more than
their expected vote share and on average governing
parties polled 88.4 percent of the vote share that
they should, using Dinkel’s method, have
‘expected’. Similar calculations can, of course, be
done for the main opposition party just as it can for
smaller parties (although Dinkel did not choose to
do this). According to the logic of the second-order
thesis, both these groups of parties should perform
better than ‘expected’ in Land elections. Figure 1
shows that the main opposition party at the federal
level does indeed perform better in the middle of
the electoral cycle, but as the next election
approaches the proportion of the expected vote that
it polls sinks again. On average, the main opposition
party registers 109.49 percent of its expected vote
and the curve that opposition party performance in
Land elections generates is a neat mirror image of
that for the federal governing parties.
Figure 2, which covers the period 1960–90,3
shows that the smaller parties – with whom voters
should theoretically be more willing to experiment
in regional elections – also perform much better in
the middle of the electoral cycle and, to some
extent, maintain their performance to the end of the
cycle when the next national election approaches.
This is due in part to the typically lower turnout in
Land elections as compared to federal elections,
which tends to magnify the impact of smaller-party
protest. On average the smaller parties poll 126.53
percent of their ‘expected vote’ and rarely dip under
the level of 100 percent. In the German case (up to
1990 at least; the post-1990 period is covered below)
there appears therefore to be solid evidence that
regional elections are indeed second-order,
subordinate to a cyclical rhythm whose scale and
tempo is set by statewide factors.
The situation in Spain is rather different. The
data in Figures 3 and 4 below are from a sample of
Spanish autonomous communities, Castilla-LaMancha,
Catalonia, Galicia and Extremadura. They
show the ‘wrong’ cyclical pattern of results. The
main party of government in Madrid achieves on
average some 87.05 percent of its ‘expected’ vote
share – less than in the German case. But, its best
European Urban and Regional Studies 2003 10(3)
202 EUROPEAN URBAN AND REGIONAL STUDIES 10(3)
0
20
40
60
80
100
120
140
160
180
0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400
Time (in days since the previous statewide election)
Percentageofexpectedvote
Hauptregierungspartei Hauptoppositionspartei
Figure 1 The results of the governing and main opposition parties at the statewide level in regional elections in Germany,
1949–90
Main governing party Main opposition party
results are precisely when they should not be: at the
middle of the statewide electoral cycle. The
performance in regional elections of the main
opposition party in Madrid also does not conform to
the expectations of the second-order model. It
achieves on average just 90.77 percent of its
expected vote, barely more than the main
government party. And again the cyclical pattern is
‘wrong’: it does worst at mid-term, when it should
be hitting peak performance. If we do the equivalent
European Urban and Regional Studies 2003 10(3)
JEFFERY & HOUGH: REGIONAL ELECTIONS 203
0
50
100
150
200
250
300
350
400
0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500
Time (in days since the last statewide election)
Percentageofexpectedvote
Figure 2 Small party performance in Land elections in Germany, 1960–90
0
20
40
60
80
100
120
140
0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300
Time (in days since the last statewide election)
Percentageofexpectedvote
Governing party Opposition party
Figure 3 The results of the main governing and the main opposition parties at the statewide level in regional elections in
Extremadura, Castilla-La-Mancha, Catalonia and Galicia, 1977–99
calculations for regionalist parties, which one might
expect in Spain to be the beneficiaries at the
statewide mid-term, we also get the ‘wrong’ curve.
It does need to be stressed that on no occasion did
regionalist parties in the four autonomous
communities get less than 100 percent of their
‘expected’ vote, and often scored twice or three
times as well as ‘expected’. Regionalist parties do
score very well in regional elections, but do best
close to and not at the mid-point between statewide
elections. Regional elections do not seem to be
subordinate to statewide elections; there seems to be
a countervailing dynamic which reverses the
expectations of the second-order thesis.
Our explanation for these differences lies in the
territorial heterogeneity of the Spanish state, which
combines a Castilian core with a number of ‘historic
nationalities’ around the periphery. Ideas about
statewide electoral cycles and regional elections as
second-order statewide elections would seem to be
less relevant in these circumstances than in
territorially more homogeneous states such as
Germany. Homogeneity manifests itself, as a rule, in
‘congruence’ of party systems at the statewide and
regional levels. In (pre-1990) Germany, with the
partial exception of the Bavarian Christian Social
Union (CSU),4 all the major parties had a statewide
‘reach’, not least because they originated as the
political expressions of non-territorial, statewide
social cleavages: class, religious belief and (in the
more recent case of the Greens) ‘materialism’ vs
‘post-materialism’. In these circumstances it was
unlikely that the terms of political debate would be
significantly differentiated by territory; voters made
decisions on Land elections in line with an
essentially undifferentiated, statewide political
debate. Regional elections were in that sense
second-order, as the ‘Dinkel curves’ suggest. In
states like Spain, where there is greater territorial
heterogeneity and where the cleavages underlying
the party system have a territorial dimension
alongside the ‘classic’ statewide cleavages of class
and religion, the subordination of regional elections
to statewide terms of debate is less likely to occur.
The failure of the Spanish ‘Dinkel curves’ to
conform to the expectations of Reif, Schmitt and
Dinkel confirms this line of thinking. Territorial
heterogeneity works against subordination to the
statewide arena; regional elections are not secondorder
and may have a status in voters’ minds
equivalent to, perhaps even higher than, statewide
elections.
These propositions receive partial support if we
look at the relationship between regional and
statewide elections in Germany since German
unification in 1990, when eastern Germany was
incorporated into the Federal Republic. After
unification Germany became a territorially much
European Urban and Regional Studies 2003 10(3)
204 EUROPEAN URBAN AND REGIONAL STUDIES 10(3)
0
50
100
150
200
250
300
350
400
0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300
Time (in days since the previous statewide election)
Percentageofexpectedvote
Figure 4 The performance of significant regional parties in regional elections in Extremadura, Castilla-La-Mancha,
Catalonia and Galicia, 1977–99
more diverse society. Eastern Germans brought
with them different patterns of socialization and a
starkly different set of economic problems and
interests into the ‘new’ Germany, opening up a new
east–west cleavage. This cleavage has found
expression in the German party system. There
remain only two genuine statewide parties – the
Christian Democratic Union/Christian Social
Union (CDU/CSU) and the Social Democratic
Party (SPD). Both the Free Democrats (FDP) and
the Greens are principally western German parties,
while the former communist Party of Democratic
Socialism (PDS) is almost exclusively a party of the
eastern Länder.
In these changed circumstances, the ‘expected’
vote-share graphs for governing, opposition and
small parties from 1990 to 2002 are substantially
different to those of the pre-unification period (see
Figures 5 and 6). In the case of the governing
parties at the federal level, the situation remains
more or less the same as before 1990. The main
governing party performs worse than ‘expected’ in
the middle of the electoral cycle and better towards
the beginning and end. So does the main opposition
party, which now seems not to get the benefit of
voters ‘punishing’ the federal government
incumbents. The smaller parties, however, have
polled unusually good results in Land elections. The
reason for this would appear to reflect the inability
of the main parties, SPD and CDU/CSU –
regardless of whether they are in the federal
government or opposition – to generate a
convincing set of messages for all voters, east and
west. Smaller parties have tapped this failure of
statewide integration by focusing protest largely
around region-specific issues. This applies to the
PDS as a defender of eastern interests; it applies to
‘flash’ parties of protest against the ‘establishment’
in Bremen and Hamburg; and it arguably applies to
the periodic successes of far-Right parties in some
Länder in mobilizing senses of alienation from
mainstream politics. We see, in other words, a
territorially more differentiated set of political
debates, not always structured by the same parties
from one Land to the next, or indeed in the same
Land when comparing statewide and regional
elections.
In sum, the proposition that regional elections
are ‘second-order’ finds, at best, only partial
confirmation in the data analysed here. An approach
that attempts to explain the relationship between the
national and regional electoral arenas in terms of the
political climate at the national level seems
insufficient, especially in cases where territorial
European Urban and Regional Studies 2003 10(3)
JEFFERY & HOUGH: REGIONAL ELECTIONS 205
0
20
40
60
80
100
120
140
160
180
0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500
Time (in days since the last statewide election)
Percentageofexpectedvote
Hauptregierungspartei Hauptoppositionspartei
Figure 5 The results of the main governing and the main opposition party at the statewide level in regional elections in
Germany, 1990–2002
Main governing party Main opposition party
cleavages are strong. In the next section we therefore
focus on this territorial dynamic, as it were reversing
the perspective on the relation of regional to
statewide elections to one taken from the ‘bottom
up’.
Multi-level voting and territorial politics
Over the last 30 years regional parties and the
institutions of regional governance have become
important components of the Western European
political landscape. But just as in the analysis of
‘orders’ of elections, the literature on territorial
politics has only rarely analysed the effects of
territorial cleavage on voting behaviour and party
competition at – and particularly between –
different levels of government. We therefore need to
re-adapt existing literatures to throw useful light on
the question of multi-level voting.
An important starting point are the cleavage
structures underlying Western European party
systems identified by Stein Rokkan, Seymour
Martin Lipset and colleagues from the late 1960s
(Lipset and Rokkan, 1967; Rokkan and Urwin,
1982; 1983; Flora et al., 1999). Among these
cleavages were conflicts of centre and periphery.
This cleavage emerged where the project of national
state building clashed uneasily with the
particularism of individual territories and peoples,
producing territorial forms of political mobilization.
This cleavage epitomized the ‘typical reactions of
peripheral regions, linguistic minorities and
culturally threatened peoples towards the uniform
and rationalising tendencies of the centralising
nation state’ (Flora et al., 1999: 282). Territorial
politics in this sense was seen as ‘bastions of
primordial local culture’, offering opposition and
resistance to the processes of national
modernization. These ‘bastions’, however, ‘rarely
survived the broadening of the electoral franchise’
(Flora et al., 1999: 283, 284). Conflicts based on the
centre–periphery divide were subsequently seen as
the residual grumblings of the losers of
modernization, as deeply parochial and as
essentially insignificant for the wider political
process. As and when territorial movements were
able to mobilize against the centre they were seen up
until the 1960s as atavistic movements disturbing
the normally peaceful flow of Western politics and
as ‘irritating anachronisms’ (Rokkan and Urwin,
1982: 1).
This understanding – even dismissal – of
European Urban and Regional Studies 2003 10(3)
206 EUROPEAN URBAN AND REGIONAL STUDIES 10(3)
0
20
40
60
80
100
120
140
160
180
0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400
Time (in days since the previous statewide election)
Percentageofexpectedvote
Figure 6 Small party performance in regional elections in Germany, 1990–2002
territorial politics was only seriously challenged in
the 1980s after a whole troupe of regional parties
had (re-)gained prominence across Western Europe.
Given their renewed vitality these parties could no
longer be written off as anachronistic phenomena
with little contemporary relevance. The analysis of
the institutions and processes of territorial politics
subsequently became a growth industry in political
science (see Urwin, 1985; Keating, 1988; 1993;
1996; De Winter, 1998; Müller-Rommel, 1998;
Hough, 2002). At the heart of this renewed interest
were three principal questions:
• Why has this new regionalism arisen? Which
combinations of cultural identity, socio-economic
structures and deficits in national politics have
facilitated the breakthrough of regional parties?
• What does the rise of new regionalism mean for
the future of the national state? Given that all
regional parties demand greater autonomy for
their territories from the centre, what are the
implications for the institutions and the
coherence of the national state?
• What effects will the new territorial politics have
on the dynamics that underpin national party
systems?
The last of these questions is naturally of most
interest for this article. Over the last 25 years or so
an important component of party and party system
analysis has been the growing volatility of
contemporary voting behaviour as well as the
processes of partisan dealignment that are held to
underlie it. The core of the dealignment thesis is
that the rise of new parties – principally Green and
far-Right – and the increasing inability of older
parties to mobilize their traditional bases of support
can be put down to an ‘uncoupling’ of social
structure and party identification. This logic can
also be applied to the breakthrough of regional
parties – although it rarely appears to have been so
applied. If parties that compete on a statewide basis
lose their integrative capacity, prospects for new,
territorial parties that articulate other interests and
sentiments open up (De Winter, 1998: 214–19;
Müller-Rommel, 1998: 25–6).
One way of capturing the effects of territorial
politics on a party system would be a measure that
could somehow reflect the weakening (or otherwise)
ability of the major parties to integrate voters at
different levels. A number of articles have thrown
light on the increasing differentiation of party
competition that appears to exist among regions,
and they have stressed that this is evidence of a rise
in the salience of centre–periphery cleavages (e.g.
Hearl et al., 1996). These contributions have
nonetheless tended to sidestep the question of
whether these ‘horizontal’ region-by-region
differences are also flanked by ‘vertical’ differences
between regional and statewide election results in
the same region. The analysis tends, in other words,
to be limited to regionally disaggregated analysis of
statewide elections. The following reason has been
put forward in justification:
As the main policy objective of ethnoregionalist parties
is the reorganisation of the national power structure
towards an increase in the degree of self-government,
and for this reorganisation only legislative bodies at the
national level are competent, ethnoregionalist parties tend
to focus on increasing their political weight at the level of
the national parliament, rather than at sub-state levels of
political representation. (De Winter, 1998: 211–12)
Such reasoning may have made sense in studying
the era in which the fight for regional autonomy is
under way, but in an era where such battles have
largely been won it is unconvincing. Belgium, Spain,
Italy, France and most recently the United Kingdom
have, over the course of the last 30 years, all created
(or in the Italian case called into practice) regional
tiers of government. It is quite conceivable that
since the granting of autonomy (in whatever form)
regional parties now choose to direct their energies
towards the newly empowered regional level of
government. And national parties also have to react
to the new circumstances, as do voters, who are now
asked to cast ballots in both the statewide and
regional arenas. Put another way – and this point
applies, too, in states like Germany or Austria that
have much longer traditions of regional elections –
do statewide and regional electoral processes (still)
follow the same logics and rhythms in the era of
regionalization?
This question takes on even more importance
when we recall the limited success of the notion of
second-order elections in explaining electoral
outcomes at the regional level in Germany and
Spain. Regional and statewide elections do not
necessarily have a hierarchical and one-way
relationship to one another. It is also conceivable
European Urban and Regional Studies 2003 10(3)
JEFFERY & HOUGH: REGIONAL ELECTIONS 207
EuropeanUrbanandRegionalStudies200310(3)
208EUROPEANURBANANDREGIONALSTUDIES10(3)
Table 1 The index of dissimilarity in selected regions (%)
Germany Spain Canada
Baden- Castilla-La- British
Bavaria Hamburg Württemb. Brandenb’g Catalonia Galicia Extremad. Mancha Columbia Ontario Manitoba Alberta
1949 19.1 7.3 16.5 6.2 33.4
1953 9.9 7.9 16.5 17.3 16.8 13.7 25.3
1957 11.6 7.7 10.2 29.8 6.3 14.6 17.3
1958 46.6 10.4 16 36.1
1961 10.3 9.2 6.6
1962 27.7 7.3 9.1 30.3
1963 28.7 13.5 5.9 35
1965 9.8 10.1 4.3 31.6 12.7 2.3 32
1968 48.7 16 18.7 51.3
1969 3.6 7.0 14.0
1972 8.5 10.7 3.1 33.6 11.1 17.1 36.9
1974 64.3 10.7 20.1 20.9
1976 3.6 12.7 4.6
1979 19.0 22.8 62 6.5 12 25
1980 4.3 12.3 8.5 59 8.5 21 24.5
1982 25.6 22.1 8.8 8.13
1983 4.0 7.2 2.5
1984 58.5 11.5 15 18.5
1986 19.4 9.6 9.0 2.74
1987 4.5 5.2 8.2
1988 41.5 14.5 4 23.5
1989 13.8 10.1 4.1 5.65
1990 4.8 8.4 13.4 8.2
1993 18.9 18.3 6.6 8.41 29 42 47 51.5
1994 4.1 22.4 7.9 11.0
1996 19.0 16.7 4.9 3.36
1997 38.5 28.5 45 52
1998 9.7 19.8 10.5 8.8
2000 14.9 6.6 7.0 12.67
2001
2002 13.9 23.6 13.5 17.9
Average 8.1 12.2 11.3 11.5 18.7 15.2 6.8 6.8 39.0 14.6 16.7 32.1
that they may have little or no relationship to each
other. In order to investigate such ideas further we
require some means of capturing whether and how
far statewide and regional elections operate
according to distinctive, even divergent, logics and
rhythms. The Western European literature remains
surprisingly unhelpful in this regard, but a sideways
glance across the North Atlantic helps to fill the gap.
Looking to Canada: the index of
dissimilarity
In Canada there have long been stark differences
between statewide and provincial election results,
and party systems are often quite different from one
province to the next. One analytical tool applied by a
number of Canadian authors in their attempts to pin
down provincial/national divergence in election
results is the calculation of an ‘index of
dissimilarity’ (Johnston, 1980; Abedi and Siaroff,
1999). Indices of dissimilarity are generated by
comparing the results of a statewide election in a
particular province with the results of the provincial
election in that province nearest in time to the
statewide election. The index expresses the
proportion of the electorate who would have to
change their vote in order for the provincial election
result to be converted into the national one. The
dissimilarity index in this way offers a clear and
straightforward measure of how divergent electoral
outcomes at different levels are. Maximum
dissimilarity would be 100 percent (all voters cast
their vote in different ways in provincial compared
to national elections in a certain province);
minimum dissimilarity would be 0 percent (no one
voted differently).
Canada is an extreme case of regional–statewide
divergence. In Canada there is no single ‘party
system’, but rather 10 provincial party systems often
with quite distinctive features, as well as one
statewide party system. The statewide party system
comprises parties that in part have only a loose
relationship with the parties that share the same
name at the regional level (Robin, 1978; Wolinetz
and Carty, forthcoming). Given this, it should not
come as a surprise that the levels of
regional–national dissimilarity are high. Voters do
routinely vote for different parties at different levels.
Parties that are strong at the one level cannot be
sure that that strength will be reflected electorally at
the other. In the most extreme example the
Conservative Party won just 1 percent of the vote in
the provincial election in British Columbia in both
1983 and 1986, but polled 47 percent of the votes in
British Columbia in the intervening federal election
of 1984! In the 1970s and 1980s approximately 60
percent of the votes that were cast in provincial
elections in British Columbia would have needed to
have been cast differently to have replicated the
result of the most recent national election. As Table
1 shows, the situation is not dissimilar in Alberta
where regional–statewide dissimilarity has long been
considerable, as it has been more recently in
Manitoba. The same applies in Ontario, particularly
since the collapse of the Conservative Party’s vote in
the early 1990s and its increasing inability to act as a
party of statewide integration. The Liberal Party is
now the only Canadian party that garners
meaningful levels of support throughout Canada in
both provincial and national elections. The Liberals
aside, the dynamics of provincial and national
elections are largely separate from one another.
Regional–statewide dissimilarity: Germany
and Spain
It is immediately obvious from Table 1 that the
statewide and regional election results for Germany
and Spain are more similar than is the case in
Canada. This is most obviously the case in
Germany. In Germany (at least before unification)
much the same set of parties competed at both the
statewide and regional levels on more or less the
same terms across the whole of the country. As a
result dissimilarity indices have rarely exceeded 10
percent. There are two periods of exception. The
first concerns the early years of the Federal
Republic. This was a period in which an initially
fragmented postwar party system quickly
consolidated into the familiar patterns that applied
through to 1990. Before party system consolidation
was complete there was considerable ‘churn’
between statewide and regional election results. The
second exception concerns the post-1990 era. As
was shown earlier, unification has introduced new
territorial dynamics to German elections. These are
European Urban and Regional Studies 2003 10(3)
JEFFERY & HOUGH: REGIONAL ELECTIONS 209
reflected also in higher dissimilarity indices, with
several results – in east and west – of over 10
percent and even 15 percent. Again we see here
evidence of the declining capacity of the main
parties to continue to integrate voters on the same
terms at both statewide and regional elections.
The Spanish data reveal a territorially uneven
pattern. In Castilla-La-Mancha and Extremadura
the dissimilarity indices are close to the lower end of
the German scale, with averages of 6.8 percent and
6.7 percent respectively over the period 1979–2000.
In Catalonia and Galicia the indices are more like
those in Canada at 18.7 percent and 16.6 percent
respectively; in these regions every fifth or sixth
voter behaves differently in regional as compared to
statewide elections. The reasons for higher
dissimilarity scores in Catalonia and Galicia would
appear to lie in the role that regionalist parties play
in the Spanish party system. In Catalonia the
strongest regionalist party, the Convergencia I Unio,
always does better in regional than statewide
elections, as does the Bloque Nacionalista Galego in
Galicia. In those regions it seems that voters make
conscious use of the possibilities of the
decentralized Spanish state to favour regionalist
solutions to political questions. In Castilla-LaMancha
and Extremadura this is not the case. These
are two of the regions in Spain that lack a strongly
defined sense of regional identity. Territory is less
important a political cleavage than one like social
class which plays out statewide. Catalonia and
Galicia are, by contrast, two of the historic
nationalities, each with a strong sense of identity,
including widely spoken national languages, and
reflected in the higher level of autonomy accorded
to them in the decentralized system. In these
circumstances there are evidently sufficient
incentives for voters to vote according to different
regional and statewide logics in regional and
statewide elections. The situation is comparable in
Canada, where language again plays a vital role in
building territorial cleavages. Add to this the sheer
diversity of regional interests and political cultures
across a vast state territory and regional and
statewide voting logics become even more dissimilar
and divergent.
Conclusions
This article set out to find ways of capturing better
the relationship of regional and statewide electoral
processes in multi-level systems. ‘Dinkel curves’ and
the dissimilarity index provide at least initial
techniques for mapping out the dynamics of multilevel
voting. Dinkel tests for the subordination of
regional elections to the electoral rhythms of
statewide politics. Dissimilarity explores how
different patterns of voting behaviour are in the
same region in regional compared with statewide
elections. Together the application of the two
techniques suggests the following conclusions.
These conclusions must necessarily remain tentative
as they draw on limited datasets from just three
country cases.
The first is that in contexts which lack deep
territorial cleavages, regional and statewide election
results are broadly similar, with political debate
structured by the same set of parties at regional as at
statewide levels, and with regional results tending to
be driven by the rhythm of a statewide electoral
cycle. Regional elections under these circumstances
are indeed second-order elections, conforming to
the assumptions that Reif and Schmitt set out. This
was the finding about pre-unification Germany.
Whether similar conclusions apply to other states
such as Austria, Australia or France which also lack
strong territorial cleavages and which have
congruent party systems at regional and statewide
levels is a question for further research.
Second, territorial heterogeneity breaks up this
pattern of subordination of regional elections. In
Germany the second-order quality of regional
elections has been challenged by the new territorial
cleavages evident since unification. There is
growing dissimilarity of regional and statewide
election results, and a reduced tendency for regional
election results to follow statewide electoral rhythms
(Hough and Jeffery, 2003b). The impact of
territorial heterogeneity is confirmed by the
Canadian and Spanish cases. Regional elections do
not conform to expectations about second-order
elections in Spain. Most interestingly, dissimilarity
indices vary considerably between regions with and
without strongly defined senses of territorial
identity. In other words, the extent to which
regional and statewide electoral processes operate
according to different logics can vary considerably
European Urban and Regional Studies 2003 10(3)
210 EUROPEAN URBAN AND REGIONAL STUDIES 10(3)
across regions within a state. States with similar
patterns of metropolitan core and distinctive
peripheries – e.g. the UK (Hough and Jeffery,
2003a) and arguably Italy – may also reveal this kind
of differentiated pattern. Canada is, it seems, an
outlier case. Regional and statewide elections seem
to have quite different logics in the minds of voters
in a state where size and territorial identities
combine to produce distinctive patterns of voting
and party systems from one region to the next.
Perhaps there are similar patterns in other big,
multinational states such as Russia or India.
In sum, the obvious analytical lens for exploring
regional elections, that of ‘second-orderness’ is
found wanting. Regional elections are evidently less
open to domination by statewide debates than, say,
European or local elections. To put it in the terms of
Reif and Schmitt, more is ‘at stake’ in regional
elections, especially where territorial cleavages
structure political debate differentially across a state.
To put it another way, the ways elections, voting
behaviour and party systems operate at the regional
level are more important a topic for political science
than has traditionally been thought.
Acknowledgements
Work for this article was carried out as part of the
research project on ‘Multi-level Electoral
Competition’ which forms part of the Research
Programme ‘Nations and Regions: the Dynamics of
Devolution’ run by the Constitution Unit,
University College London and funded by the
Leverhulme Trust.
Notes
1 Though Dinkel does factor in the supplementary
influence of regional party system factors and Land
government incumbency, he consciously excludes
potential differential effects produced by different
incumbent parties at the federal level, ‘events’ with
federal or Land electoral resonance, socio-structural
differences between the Länder, personalities and the
economic cycle. See Dinkel (1977: 351–2).
2 Though a review of the data Dinkel analysed shows that a
number of his entries were miscalculated.
3 The period 1949–60 was one in which a fragmented
multi-party system consolidated into a simple three-party
system, in part due to adjustments to the electoral system.
It is therefore an outlier period which we leave to one side.
4 This is a partial exception. The CSU forms a single
parliamentary fraction in the federal parliament with the
Christian Democratic Union. The CDU stands for
election across all Germany except Bavaria; the CSU does
not put up candidates for election outside Bavaria.
CDU/CSU together are to all extents and purposes a
common party at federal level.
References
Abedi, A. and Siaroff, A. (1999) ‘The Mirror has Broken:
Increasing Divergence Between National and Land
Elections in Austria’, German Politics 8 (1): 207–28.
De Winter, L. (1998) ‘Conclusion: a Comparative Analysis
of the Electoral, Office and Policy Success of
Ethnoregionalist Parties’, in L. De Winter and H.
Türsan (eds) Regionalist Parties in Western Europe, pp.
204–47. London: Routledge.
Dinkel, R. (1977) ‘Der Zusammenhang zwischen Bundesund
Landtagswahlergebnissen’, Politische
Vierteljahresschrift 18 (3): 348–60.
Flora, P., Kuhnle, S. and Urwin, D. (1999) State
Formation, Nation-building and Mass Politics in Europe:
the Theory of Stein Rokkan. Oxford: Oxford University
Press.
Hearl, D., Budge, I. and Peterson, B. (1996)
‘Distinctiveness of Regional Voting: a Comparative
Analysis Across the European Community Countries
(1979–93)’, Electoral Studies 15 (1): 167–82.
Heath, A., McLean, I., Taylor, B. and Curtice, J. (1999)
‘Between First and Second Order: a Comparison of
Voting Behaviour in European and Local Elections in
Britain’, European Journal of Political Research 35 (3):
389–414.
Hough, D. (2002) The Fall and Rise of the PDS in Eastern
Germany. Birmingham: Birmingham University Press.
Hough, D. and Jeffery, C. (2003a) ‘Elections in Multi-level
Systems: Lessons for the UK from Abroad’, in R.
Hazell (ed.) The State of the Nations 2003. Thorverton:
Imprint.
Hough, D. and Jeffery, C. (2003b) ‘Landtagswahlen:
Bundestestwahlen oder Regionalwahlen’, Zeitschrift für
Parlamentsfragen 32 (1): 79–94.
Jeffery, C. and Hough, D. (2001) ‘The Electoral Cycle and
Multi-level Voting in Germany’, German Politics 10 (2):
73–98.
Johnston, R. (1980): ‘Federal and Provincial Voting:
Contemporary Patterns and Historical Evolution’, in D.
European Urban and Regional Studies 2003 10(3)
JEFFERY & HOUGH: REGIONAL ELECTIONS 211
J. Elkins and R. Simeon (eds) Small Worlds: Provinces
and Parties in Canadian Political Life, pp. 106–30.
Agincourt: Methuen.
Keating, M. (1988) State and Regional Nationalism –
Territorial Politics and the European State. London:
Harvester Wheatsheaf.
Keating, M. (1993) The Politics of Modern Europe.
Aldershot: Edward Elgar.
Keating, M. (1996) Nations against the State. New York: St
Martin’s Press.
Lipset, S.M. and Rokkan, S. (1967) Party Systems and
Voter Alignments. New York: Free Press.
McLean, I., Heath, A. and Taylor, B. (1996) ‘Were the
1994 Euro- and Local Elections in Britain Really
Second Order? Evidence from the British Election
Panel Study’, in D.M. Farrell, (eds) British Elections and
Parties Yearbook, pp. 1–20. London: Frank Cass.
Müller-Rommel, F. (1998) ‘Ethnoregionalist Parties in
Western Europe: Theoretical Considerations and
Framework of Analysis’, in L. De Winter and H.
Türsan (eds) Regionalist Parties in Western Europe, pp.
17–27. London: Routledge.
Reif, K. (1985): ‘Ten Second-order Elections’, in K. Reif
(ed.) Ten Second-order Elections, pp. 1–36. Aldershot:
Gower.
Reif, K. and Schmitt, H. (1980) ‘Nine Second-order
National Elections: a Conceptual Framework for the
Analysis of European Election Results’, European
Journal of Political Research 8 (1): 3–44.
Robin, M. (1978) Canadian Provincial Politics. London:
Prentice Hall.
Rokkan, S. and Urwin, D.W. (1982) The Politics of
Territorial Identity. London: Sage.
Rokkan, S. and Urwin, D. (1983) Economy, Territory and
Identity: Politics of West European Peripheries. London:
Sage.
Urwin, D.W. (1985) ‘The Price of a Kingdom: Territory,
Identity and the Centre–Periphery Dimension in
Western Europe’, in Y. Meny and V. Wright (eds)
Centre–Periphery Relations in Western Europe, pp.
151–70. London: George Allen & Unwin.
Van der Eijk, C., Franklin, M. and Marsh, M. (1996)
‘What Voters Teach Us About Europe-wide Elections:
What Europe-wide Elections Teach Us About Voters’,
in Electoral Studies 15 (2): 149–66.
Wolinetz, S. and Carty, K. (forthcoming) ‘Multi-level
Electoral Competition: the Canadian Case’, in C. Jeffery
and D. Hough (eds) Devolution and Electoral Politics:
Lessons from the European Experience. Manchester:
Manchester University Press.
Correspondence to:
Dr Dan Hough, School of Politics, University of
Nottingham, University Park, Nottingham
NG7 2RD, UK.
[email: daniel.hough@nottingham.ac.uk]
European Urban and Regional Studies 2003 10(3)
212 EUROPEAN URBAN AND REGIONAL STUDIES 10(3)