The geographical dynamics of Italian electoral change,
1987e2001
Michael E. ShinÃ, John Agnew
University of California at Los Angeles, Department of Geography, 1255 Bunche Hall, Los Angeles, CA 90095-1524, USA
Abstract
Much has been written about why the old Italian party system collapsed in the early 1990s, and the various features of the one
that replaced it, such as the mixed electoral system and the emergence of the regionalist party, the Northern League. Relatively little
research attempts to show how old parties were replaced as the system collapsed from 1987 to 1994, and as the subsequent system
emerged and consolidated between 1992 and 2001. This paper uses spatial analysis to examine the geographical pattern of support
for the fading and rising parties to show how the old Italian party system was replaced. Instead of seeing the geography of voting as
a reflection of underlying social cleavages, on the assumption that new parties would just slot into the electoral ‘‘spaces’’ of the old
ones, we see it as a useful diagnostic that can inform how one set of parties is replaced by another.
Ó 2006 Elsevier Ltd. All rights reserved.
Keywords: Electoral change; Party replacement; Spatial analysis; Italy
1. Introduction
Across elections there is always a flow of votes from
one party to another, and into and out of the electorate.
In voting studies there is a large literature that addresses
this very question of electoral change from one election
to the next (e.g., Burnham, 1970; Butler and Stokes,
1969; Caciagli and Spreafico, 1990; Katz, 1973; Key,
1955). For the most part, this body of work is premised
upon two key assumptions. The first assumption is that
political change usually occurs through the ideological
mutation of political parties, and the second related assumption
is that party representation remains constant
between most elections. In other words, it is presumed
that support for parties will ebb and flow as their positions
on various issues change, and that existing parties
that appear on the ballot in one election will also appear
on the ballot in subsequent elections, ad infinitum.
These assumptions may not be unreasonable in majoritarian
electoral systems with two dominant parties, but
it is certainly problematic theoretically and in relation
to more complex electoral and party systems. For instance,
what happens when there is a period of partysystem
change or in new democracies where party
politics has yet to consolidate? Party systems usually
disappear only with the onset of a period of dictatorship
or as a result of dramatic shocks such as wars, and
when they reemerge it is often in a new form as if
they had been remade from scratch. After World War
II countries such as Japan, France, Italy and West Germany
all went through this process. Similarly, many
central and eastern European countries recently
à Corresponding author. Tel.: þ1 310 825 1071.
E-mail addresses: shinm@geog.ucla.edu (M.E. Shin), jagnew@
geog.ucla.edu (J. Agnew).
0261-3794/$ - see front matter Ó 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.electstud.2006.05.002
Electoral Studies 26 (2007) 287e302
www.elsevier.com/locate/electstud
experienced the rebirth of party politics as part of their
transitions from single-party rule (Kitschelt, 1992,
1995), and there are signs that fledgling party systems
are developing in places like Afghanistan and Iraq.
As in central and eastern Europe so in Italy, the end
of the cold war, which had divided post-World War II
Italian politics between a dominant Christian Democratic
party and the largest Communist party in the
free world, combined with a corruption scandal involving
the major parties of government, to precipitate the
collapse of the Italian party system in the early 1990s.
This collapse and the subsequent reemergence of the
so-called ‘second Italian republic’ is well documented
(e.g., Bartolini and D’Alimonte, 1995; D’Alimonte
and Bartolini, 2002; Diamanti and Mannheimer, 1994;
Gangemi and Riccamboni, 1997; Gundle and Parker,
1996). Most attention, however, has been given to either
why the old system collapsed (the end of the Cold War,
the activism of the judiciary in Milan in prosecuting
corruption by the old parties, the breakdown of the system
of sharing the spoils of government, etc.) or to specific
features of what followed: the new electoral system
of 1993 (e.g., D’Alimonte and Chiaramonte, 1995; Pappalardo,
1995); the role of Italy’s wealthiest businessman,
Silvio Berlusconi, in the new party system (e.g.,
McCarthy, 1996); the difficulty of creating a united
left from left-Catholic and Communist fragments
(e.g., Cartocci, 1994; Ignazi, 1992); the disruptive character
of Umberto Bossi’s Northern League (e.g., Diamanti,
1993, 1996); and the rehabilitation of the far
right in Italian politics (e.g., Tarchi, 1997). Relatively
little research shows how new political parties replaced
old ones as the old system collapsed from 1987 to 1994
or how the new party system has consolidated since this
dramatic period in Italian politics (but see Wellhofer,
2001).
This paper uses spatial analysis to examine the geographical
patterning of support for the fading and rising
parties in order to show how the Italian party system
was replaced and/or reconsolidated between 1987 and
2001. Rather than seeing the geography of electoral
support simply as a reflection or mapping of underlying
social cleavages or transcendental issues (capital vs. labor,
left vs. right, protectionist vs. free trade, etc.) in the
electorate, on the assumption that the new parties would
just slot into the electoral ‘‘spaces’’ of the old ones, the
geography of voting is considered to be diagnostic of
the several socio-spatial processes by which one set of
parties replaced another. More precisely, we examine
whether replacement is: (1) the straightforward substitution
or switching of old parties with new ones; (2)
the splitting of old parties among new parties; (3) the
colonizing of areas by a new party from adjacent areas
where the old one was weak; or (4) the mobilizing of
new parties because they represent issues and ideologies
that were not previously represented. The next section
frames this question about how political change
occurs within the existing body of work on the subject,
and is followed by a brief review of the recent crisis of
Italian politics. This sets the stage for the spatial analysis
that serves to sharpen thinking about processes of
political change and to illustrate the utility of a particular
geographical framework of analysis for electoral
studies.
2. Political change and party replacement,
Italian style
Studies of electoral change tend to disregard how it
occurs with different parties, different people and in different
places, in favor of explaining why change occurred,
often presuming that how it happens is
geographically invariant. This presumption is often incorrect.
For instance, Nardulli (1995) uses US presidential
election data to illustrate clearly how the process of
realignment in America was played out differently in
different regions during the twentieth century. Other
electoral studies also show that there are definite geographies
to swing voting and ticket-splitting in places
such as Germany (Gschwend et al., 2003), New Zealand
(Johnston and Pattie, 1999) and Britain (Pattie and
Johnston, 2003). By recognizing political change to
be a geographically contingent process, further insights
into both how and why such changes occur can arguably
be obtained.
A period of party-system change, like that experienced
by Italy in the early 1990s, is quite useful for empirically
addressing how electoral change takes place
because what might seem a reasonable assumption
about the uniformity and spatial homogeneity of electoral
swing in normal periods is now demonstrably
open to doubt. The disruption of established partisan
identifications allows for greater potential variance in
the range of processes involved in how electoral change
occurs. How votes flow from party-to-party and in and
out of the electorate can no longer be presumed as geographically
uniform and, analogously, as reflecting
a single calculus across all voters. Answering the question
why electoral change takes place, therefore, needs
a more complex understanding of how it occurs that is
aided by examining disruptive periods such as that in
Italy between 1987 and 2001.
Periods of dramatic party system change make them
ideal for examining how electoral change takes place
288 M.E. Shin, J. Agnew / Electoral Studies 26 (2007) 287e302
for three reasons. First, the menu of political choice is
no longer constant, so the ‘‘swing’’ from one party to
another takes on a different meaning when the old party
no longer exists or, even if still offering scattered candidates,
is totally eclipsed by one or several new parties.
Second, new parties may not necessarily represent the
same cleavages, divisions, or issues upon which support
for the old parties rested, so it cannot be presumed that
replacement simply involves switching from a single
old party to a single new one. Third and finally, new
parties might concentrate their electoral efforts in different
places from the old ones by ‘‘specializing’’ regionally
or picking certain targeted demographic,
social, and cultural constituencies for particular attention,
including groups hitherto shunned or ignored.
Though there is a tradition of research on party organization
and party change (e.g., Janda, 1980; Katz and
Mair, 1994; Panebianco, 1988), relatively little has
been written on the specific outcomes of transitions between
party systems and, more specifically, on how
such transitions are navigated in electoral terms. Such
periods are relatively rare, particularly when wars and
periods of dictatorship are excluded, but they occur
and there are analogies between them and situations
of dramatic electoral change involving third parties
and large vote swings between previously dominant
parties. The ‘remaking’ of existing major parties in
the US and UK has also biased attention away from
cases where new parties spring up to replace long established
ones. Moreover, the cleavage and rational choice
models of political behavior would both predict a tendency
for replacement to be synonymous with substitution.
Both perspectives see political expectations in
terms of relatively fixed political-issue ‘‘spaces’’ in
which the various dimensions of political choice (left/
right, Catholic/anticlerical, etc.) are given. Consequently,
new parties will tend to slot into electoral space
more or less as substitutes for the ones that have left.
This is a problematic presumption because party
replacement is not exhausted by substitution.
2.1. Types of party replacement
The primary theoretical question of this paper concerns
how the term ‘‘replacement’’ is understood in
studies of political change. One party replacing another
one in electoral competition can be thought of in terms
of a number of understandings of the term replacement.
In one meaning replacement is akin to substitution in
which the new party takes over all or a large share of
the votes of a previous one that has disappeared. This
is the quantitative electoral sense in which the term is
used by most commentators. In another related understanding,
however, replacement involves a number of
parties splitting the votes of an old one among them.
In a third scenario the collapse or disappearance of
one party draws attention to a crisis in the social world
that the old party had long represented. This allows an
existing, opportunistic party to move in and colonize
that world, not necessarily by taking over the old vote
so much as providing a voice for and mobilizing new
voters and elements increasingly alienated from the
old party even before its final disappearance.
This third understanding of the term ‘‘replacement’’
draws particular attention to three important theoretical
premises when thinking about how new parties achieve
support as old ones disappear. These premises are geographical.
This is why understanding replacement
means understanding its geographical forms. One
premise is that electoral choices should be seen in relation
to the discrete social-territorial settings or places in
which such choices are exercised. Old parties do not
simply disappear without reason even when they ‘‘rot
from the head down’’, i.e., disintegrate organizationally
before they disappear in electoral terms. They are often
no longer in tune with local social mores, and their electoral
persistence is more a matter of inertia than persisting
enthusiasm from a popular base. New issues such as
the environment, foreign immigration, globalization,
and women’s rights create new constituencies and divide
populations in new ways. But choices made also
reflect the choices available. These are a function of
candidate lists, the strength of party organization, campaign
strategies, and local issues feeding into ones of
wider relevance upon which parties run for election.
Second, parties are not simply electoral vehicles, although
orthodox thinking in Anglo-American political
science often regards them as such, perhaps because
the American parties (as opposed to individual representatives)
often seem unrelated to particular population
or sectional interests. Parties can be more or less
effective intermediaries between state and society,
channeling resources from center to periphery and rewarding
some social and territorial interests at the expense
of others. In this sense the distinction between
mass and patronage parties is a false dichotomy: all
parties are patronage parties. Judgments are made about
how effective the party is in ‘‘delivering the goods’’ and
whether or not we (in our place) are being rewarded
more or less than they (in their place). Much of the geography
of party politics is a result of who gets what,
when, and where more than a reflection of ‘‘underlying’’
or foundational social cleavages that have a
geographical bias to them.
289M.E. Shin, J. Agnew / Electoral Studies 26 (2007) 287e302
Third, new parties can have totally different symbolic,
interest and strategy repertoires from the parties
they replace. In particular, they can appeal to new territorial
formulations of the dilemmas that the old party
dealt with through allocations of public resources
(Agnew, 1997). Ethnic and regionalist parties are the
most obvious exponents of such a territorialized approach.
Typically they focus on regional relative deprivation
or a sense of resentment at the relative success
of other regions in commanding state resources or acquiring
more than an average per capita share of national
revenues. But nationalist parties wanting to
expel foreigners and put up protectionist barriers and
liberal parties desiring to remove all limits on trade
and investment are also engaged in territorial
reframing.
Since party replacement is underpinned geographically,
a spatial analytic approach is necessary to disentangle
the types of replacement identified above. Votes
are indeed channeled between parties from one election
to the next, and though it may be convenient to view
such electoral change as geographically invariant,
such flows are seldom uniform across any democracy.
Therefore, in order to completely understand electoral
change it is necessary to elucidate how it occurs. Before
examining this process in Italy, a brief overview of the
recent Italian political crisis is provided and the parties
of interest in this analysis are identified.
2.2. Party crisis and party change in Italy
The old regime of parties that had dominated Italy
since 1947 collapsed between 1987 and 1996. During
the old regime, the main parties of Italy campaigned nationally
but each had regional bases in which their support
was consistent and concentrated (e.g., Agnew,
1987b; Galli and Prandi, 1970; Shin, 2001; Trigilia,
1986). The Christian Democrats (DC) found support
in the northeast (the ‘white’ zone) and the south, and
support for the Italian Communist Party (PCI) was
strongest in the center of Italy (the ‘red’ zone). Only
during the period 1963e1976 did the DC and PCI, the
two largest parties in Italy, achieve a significant degree
of nationalization in their levels of popular support (Agnew,
2002). Other parties remained limited to particular
regions (e.g., the fascist Social Movement for Italy
(MSI) in the urban south) or switched regional bases
in terms of their highest levels of support (e.g., the
core base of support for the Italian Socialist Party
(PSI) moved south from its original northwest base)
(Agnew, 1988). Smaller parties that were situated along
the center of the political spectrum tended to be
clientelistic and relied on local ‘notables’ for their electoral
success.
The end of communism across Europe served to accelerate
and exacerbate the process of party change in
Italy. In particular, shortly after the fall of the Berlin
Wall, the leadership of the PCI proposed self-reform
in order for the party to advance within parliament
(see Baccetti, 1997; Hellman, 1992; Ignazi, 1992; Kertzer,
1996). Support for the initiative to reform the PCI
into the Democratic Party of the Left (PDS), however,
was not unanimous. A small group opposed to discarding
communist ideology, which also drew support from
the radical Proletarian Democracy (DP) party among
other left-wing groups, splintered off to form the Communist
Refoundation party (RC) (Foot, 1996). Hence,
in 1991, the left of the Italian political spectrum that
was once largely occupied and dominated by the PCI,
was now split between the RC and PDS.
The reconfiguration of the Italian left and the end of
the cold war also removed the perceived ‘red’ threat that
the PCI had come to embody during its four decades of
exclusion from government. This was significant because
the success of the ruling party, the Christian Democrats,
was in large part based upon its anticommunist
stance. The disappearance of the PCI effectively freed
the Italian electorate to vote as it liked, and to express
its disgust and frustration with the DC’s underachievement
and widespread corruption, the extent and beneficiaries
of which were revealed publicly by the Mani
pulite (‘Clean Hands’) and Tangentopoli (‘Bribesville’)
investigations of the early-1990s (see McCarthy, 1997;
Nelken, 1996). The disintegration of the DC in 1993 is
viewed as an ‘‘anticipated reaction’’ to the negative fallout
related to the investigations into government corruption,
and also resulted in the disappearance of the
DC’s partner in crime, the Italian Socialist Party (PSI)
(Morlino, 1996, 9). As notable party leaders, such as
the PSI’s Bettino Craxi, came under public scrutiny,
others defected to create smaller parties, but a sizeable
political vacuum nevertheless remained for the taking in
the center of Italian politics.
The meteoric rise of Silvio Berlusconi’s Forza Italia
into this political void was unlike anything witnessed
previously in Italian politics. Using the chant for the national
soccer team as the party name, the success of
Forza Italia can be attributed to it appearing with the
right message at the right time, and to Berlusconi’s
use of his three privately owned television networks
before the 1994 elections (McCarthy, 1996). Perhaps
the most interesting political party to emerge during
this period was the regionalist Northern League
(LN). Comprised of various regional and local political
290 M.E. Shin, J. Agnew / Electoral Studies 26 (2007) 287e302
organizations located in the northern regions of the peninsula,
the Northern League and its outspoken leader,
Umberto Bossi, has used particular territorial identities
as their main rallying point. Over time, the language of
the League has shifted to and fro between emphasizing
social, economic and cultural differences between the
prosperous north and public expenditure-hungry and
corrupt south to the outright secession of the north
which Bossi refers to as Padania (see Agnew, 1995; Diamanti,
1993, 1996). Also emerging out of the political
turmoil of the early-1990s was the post-fascist National
Alliance (AN) led by Gianfranco Fini. By strategically
softening its fascist overtones, and joining forces with
the more conservative elements of the defunct DC, the
AN promoted a more democratic image and has since
become an important component of Berlusconi’s center-right
coalition.
3. Spatial analysis of party replacement in Italy
In light of the crisis of parties in Italy, a technique
that is sensitive to the processes underlying party replacement,
and that recognizes that electoral change is
not regulated by a singular logic that can be applied arbitrarily
across all territorial units in the same fashion, is
particularly useful. In certain circumstances, there may
be no direct relationship between old and new parties, in
other situations one or more parties may be the direct
descendents of a disappeared party and still other cases
may reveal that electoral flows to a new party tend to
vary considerably across Italy. The ability to distinguish
between different types of party replacement (i.e., colonization/mobilization,
vote splitting and substitution)
and levels of political change in Italy indicates the
need to frame politics geographically.
In order to differentiate between the various types of
party replacement, and to evaluate political change in
Italy, a spatial analytic framework is used. Spatial analysis
is premised upon the fact that spatially referenced
data are ‘‘special’’ (Anselin, 1989; Haining, 2003;
O’Sullivan and Unwin, 2003). Spatially or geographically
referenced data refer to data that are collected
on the basis of identifiable locations or places, such as
census enumeration units, electoral precincts or in this
case Italian provinces and comuni or municipalities.
What makes spatial data special is that they tend to be
biased by the locations and places whence data are
collected, and frequently, values of individual observations
are correlated with each other and are thus not independent
as classical probability theory necessitates.
The geographic concentration or local clustering of
similar values in a data set is formally called spatial
autocorrelation or spatial dependence. The presence
of spatial dependence, for example in the dependent
variable of a regression equation, may result in biased
and inconsistent parameter estimates, and subsequent
incorrect inferences (Anselin, 1988).
Within electoral studies, spatial analysis and geographic
methodologies are gaining recognition as useful,
if not always necessary, techniques (e.g., Calvo
and Escolar, 2003; Kim et al., 2003). As more and
more geo-referenced election data become available,
researchers are increasingly aware of both the opportunities
and implications that come with working with
spatial data. For instance, theories of voting that consider
socio-geographic processes (e.g., diffusion of politically
relevant information, neighborhood effects)
and place-based structures (e.g., social networks, institutions)
that can bias political information and behavior
can benefit from spatial analytic methods. O’Loughlin
et al. (1994) used spatial analysis to show how Nazi support
in Weimar Germany was geographically contingent,
and similar procedures have been used to
examine parliamentary elections in Russia (O’Loughlin
et al., 1996). In the United States context, Cho (2003)
has examined ethnic campaign contribution networks
with spatial analysis, suggesting that processes of diffusion
and contagion are likely to be at work. Such geographic
perspectives on voting behavior are not
without their critics (see King, 1996; McAllister,
1987), but the nature of georeferenced data and insights
provided by spatial analytic methods illustrate the need
at least to consider geography more than as a mere
backdrop when examining electoral behavior (Agnew,
1996; Johnston, 1992). In addition to the detection
and evaluation of spatial autocorrelation in Italian election
data, we show how spatial analysis can be used to
examine electoral change, and to disentangle and clarify
how party replacement is played out across the
country.
The following spatial analyses use election returns
from the proportional representation (PR) component
of the 1987, 1992, 1994, 1996 and 2001 Italian national
elections. A mixed PR/majoritarian electoral system
was introduced in 1993, and was subsequently abolished
in October 2005 in favor of a return to a PR. Since
our particular interest is in how old parties were replaced,
and not necessarily voter transitions between
concurrent elections (e.g., see Wellhofer, 2001), we
are restricted to using the PR element because it permits
comparability between the old (pre-1993) and new
(post-1993) party systems. The data, provided by the
Cattaneo Institute in Bologna and reported at the level
of the Italian municipality (N > 8000), were aggregated
291M.E. Shin, J. Agnew / Electoral Studies 26 (2007) 287e302
up to the level of Italian provinces. Eight new provinces
have been introduced over the course of the study period,
but to facilitate consistent spatial and temporal
comparisons, the 1987 Italian provincial configuration
is maintained for the entire analysis. Note also that
the region of Val d’Aosta is excluded from the analysis
because it is a single member constituency and does not
elect a deputy using PR.
The spatial analysis of party replacement in Italy
concentrates on how old parties, and in particular the
Christian Democrats (DC), the Italian Communists
(PCI) and the neo-fascist Social Movement for Italy
(MSI) were replaced. We focus on these parties for three
reasons. First, the political parties that arguably have replaced
the DC, PCI and MSI are among the most important
post-crisis political actors in Italy. Though the
Italian Socialists could be included in the old party
list, the decision to include the MSI is based upon an attempt
to cover the breadth of the Italian political spectrum.
Table 1 provides a comparative overview of
support for old and new parties between 1987 and
2001. The vertical arrangement of parties mimics the
span of the Italian political spectrum with left-wing
parties placed in the topmost rows, centrist parties in
the middle and right-wing parties in the bottom rows.
The ‘Other’ category aggregates the support for parties
across the political spectrum that received less than 5%
of the total vote.
The second reason we focus on these parties is because
the respective motivations and circumstances surrounding
the disappearance of the PCI, DC and MSI,
respectively, were fundamentally different. Hence,
there is reason to believe that the processes behind replacement
were different for each party. Detailed examinations
of how the PCI, DC and MSI were replaced
illustrate whether the concept of replacement is
equivalent to substitution or whether other understandings
need to be considered. Finally, the geographies of
support for the DC, PCI and MSI were regionally concentrated
and more or less complemented one other.
Fig. 1 maps provincial levels of support for each party
at the last time each appeared on the ballot.
Determining whether or not, and to what extent, the
geography of support for the new replacement parties
overlap these historical strongholds illustrates how the
process of party replacement may be geographically
contingent in certain cases. It is well established that
levels of electoral support for the new and the old political
parties in Italy exhibit varying degrees of local clustering
as well as regional diversity (see Agnew et al.,
2002; Cartocci, 1990; Diamanti and Mannheimer,
1994; Galli and Prandi, 1970; Shin and Agnew, 2002).
These geographic clusters and variations are clearly evident
in Fig. 1, and can in part be attributed to spatial
processes that rely upon the interdependence of places
(Haining, 2003). Spatial processes such as diffusion, exchange
and transfer, interaction and dispersal are frequently
manifested as spatial autocorrelation in
quantitative data, which at the local scale may reflect information
networks, the spatial (re)-organization of political
parties or any number of social, economic,
cultural and/or political interactions or exchanges.
The processes that contribute to such local clustering
also account for the geographic variation in electoral
support and in changes at the regional and national
levels (e.g., Johnston and Pattie, 1992; Pattie and Johnston,
1997). The theoretical argument underlying such
regional variations is that voters mediate and respond
to myriad social, economic and political processes
that influence political attitudes and behavior differently
in different places (Agnew, 1987a; Johnston,
1986, 1992).
Exploration of the spatial structure of party replacement
in Italy begins with the formal detection of spatial
autocorrelation of electoral support for the major parties
in competition between 1987 and 2001. The classic indicator
used to identify the presence of spatial autocorrelation
is Moran’s I (e.g., see Cliff and Ord, 1981;
Griffith, 1987). Moran’s I can be expressed as:
I ¼
P
i
P
j wijðxi À xÞ
À
xj À x
Á
P
iðxi À xÞ2
where w is an element in a row-standardized spatial
weights matrix, W, that summarizes the geographic
linkages between observations (i.e., 1 ¼ adjacency;
0 ¼ separation), x is the vote share for the party of interest,
and x is the average vote share for the party of
Table 1
Vote shares for Italian political parties, 1987e2001
1987 1992 1994 1996 2001
Italian Communist Party (PCI) 26.6
Communist Refoundation (RC) 5.6 6.0 8.6 5.0
Democratic Party of the
Left (PDS)
16.1 20.4 21.1 16.3
The Daisy Party (MRG) 14.3
Italian Socialist Party (PSI) 14.3 13.6
Christian Democrats (DC) 34.3 29.7
Forza Italia (FI) 21.0 20.6 29.1
Northern League (LN) 8.7 8.4 10.1 4.1
National Alliance (AN) 13.5 15.7 11.9
Italian Social Movement
(MSI)
5.9 5.4
Other 18.9 21.0 30.7 24.0 19.2
292 M.E. Shin, J. Agnew / Electoral Studies 26 (2007) 287e302
interest. Contiguity or adjacency is used to capture the
geographic relations between provinces in the spatial
weights matrix W. Other methods exist to fill such
weights matrixes (e.g., selection of a cut-off distance),
but contiguity is useful here because it corresponds to
the minimum distance between provinces, and serves
as a good surrogate for local-level interdependencies.
Moran’s I is a numeric index that describes the nature
of the relationship between each observation and
its neighbors. One way to compare the value of a location
to its neighbors is to calculate a spatially lagged
variable (Anselin, 1992). The spatial lag is in effect
the weighted average of values that surround any given
observation (i.e.,
P
j wijðxj À xÞ). Used in many statistical
indicators of spatial association, spatially lagged
variables provide a simple way to capture and to summarize
geographic relationships in quantitative data.
A statistically significant, positive value of Moran’s I indicates
the presence of spatial dependence, and a significant
negative value suggests a pattern of dissimilarity.
To evaluate the significance of Moran’s I the standard
deviate or z-score of the index is used based upon the
assumption that values of x are distributed randomly
in space (for details, see Anselin, 1988; Cliff and Ord,
1981; Griffith, 1987). Moran’s I values for the parties
of interest in the new and old party system are reported
in Table 2.
Values of Moran’s I for all parties are positive and
highly significant, indicating that electoral support
for each party of interest is not distributed randomly
across the Italian peninsula. When comparing levels
of clustering between old and new parties, or over
time, both continuities and variations are identifiable.
For example, there is a certain degree of continuity,
at least initially, in the value of Moran’s I for parties
of the left (i.e., PCI, [P]DS, RC and MRG), but with
each subsequent election, values decrease and move
further away from the value associated with the PCI.
Similar trends exist for the Christian Democrats (DC)
and what are considered its primary replacements,
Forza Italia (FI) and the Northern League (LN).
Finally, support for the post-fascist National Alliance
(AN) appears to be more clustered than for its predecessor
the Italian Social Movement (MSI). Notwithstanding
assumptions, correct or incorrect, about
which new parties replaced old ones, electoral support
for all parties of interest is significantly clustered in
each election.
To obtain more detailed insights into how party replacement
was played out across Italy, a local indicator
of spatial association (LISA), and in particular, the local
< 20%
20% - 26%
27% - 33%
> 33%
< 23%
23% - 31%
31% - 38%
> 38%
PCI 1987 DC 1992 MSI 1992
< 4%
4% - 5%
5% - 7%
> 7%
Fig. 1. Electoral support for the PCI in 1987, DC in 1992 and MSI in 1992.
Table 2
Moran’s I scores for selected old and new parties, 1987e2001
1987 1992 1994 1996 2001
Old party system
PCI 0.702
DC 0.525 0.706
MSI 0.425 0.496
New party system
(P)DS 0.730 0.721 0.720 0.690
RC 0.702 0.798 0.597 0.503
MRG 0.246
FI 0.743 0.577 0.374
LN 0.750 0.716 0.715 0.590
AN 0.796 0.692 0.583
All values are significant at the p < 0.001 level.
293M.E. Shin, J. Agnew / Electoral Studies 26 (2007) 287e302
Moran statistic, is used. The equation for the local Moran’s
I index is:
Ii ¼
zi
P
z2
i
X
j
wijzj
where z is measured in deviations from the mean and inference
is based upon a conditional randomization
method (see Anselin, 1995a,b). Unlike the global Moran’s
I index which returns a single value for an entire
data set, local Moran values are calculated for each individual
observation, or in this case, each Italian province.
Local Moran’s I statistics provide information
about the degree and nature of clustering around each
observation by determining the ‘contribution’ that
each observation makes to the overall global statistic.
In fact, the average of local Moran’s I statistics is equivalent
to the global Moran’s I value. Since a local Moran’s
I index is calculated for each observation,
significant values (i.e., p < 0.05) can be mapped to
identify positive spatial dependence (i.e., high values
surrounded by similarly high values, or low values surrounded
by similarly low values), or spatial outliers
(i.e., high values surrounded by low values, or low
values surrounded by high values).
To complement the local Moran’s I statistic, the bivariate
version of the local Moran is invoked. The equation
for the bivariate LISA is identical to the local
Moran, but a different variable is used for the spatial
lag component (e.g.,
P
j wijyj). In this analysis, the spatial
lag component is the respective mean deviated level
of support for the old parties of interest (i.e., MSI, PCI,
DC), the last time that each appeared on the ballot.
Hence, the bivariate LISA serves as a diagnostic for
evaluating the geography of party replacement. More
precisely, the bivariate LISA permits the identification
of four types of spatiotemporal correlation:
I. high levels of support for new party z surrounded
by high levels of support for old party y; is consistent
with switching and possibly the splitting
of votes between two or more parties;
II. low levels of support for new party z surrounded
by high levels of support for old party y; suggests
the new party is unable to colonize areas where
old party y once enjoyed success;
III. low levels of support for new party z surrounded
by low levels of support for party y; is consistent
with a certain degree of spatiotemporal stability,
like type I above;
IV. high levels of support for party z surrounded by
low levels of support for party y, suggests that
the new party may be mobilizing a new segment
of voters or colonizing areas where the old party
was weak.
Given the overview of the recent crisis of the Italian
party system, and the fact that electoral support for most
parties in both the old and new Italian party system is
spatially dependent (see Table 2), there is reason to expect
that spatiotemporal relations exist between votes
for the old and the new parties. Conversely, since replacement
arguably is not a singular process, or one
that is completely explained by direct substitution, an
approach that can elucidate how replacement is played
out across the Italian peninsula, in different locations, is
constructive. Mapping and comparing LISA and bivariate
LISA statistics between elections permits us to locate
areas and regions of spatiotemporal stability and
instability, which in turn reveal the possible sociogeographic
processes that underlie party replacement
in Italy. Such a technique is also quite useful, if not necessary,
when calculating flow-of-vote estimates or when
estimating voter transitions between parties. For instance,
one of the assumptions underlying the implementation
of Gary King’s (1997) ecological inference
approach, which can be used to estimate voter transitions,
is the absence of spatial autocorrelation. Exploratory
spatial data analyses, such as those described
previously, can be used to evaluate whether or not spatial
dependence exists within a dataset, and to identify
areas where spatial dependence is relatively uniform
and can be considered accordingly.
3.1. Replacement as substitution: from MSI to AN
As noted earlier, party replacement is most frequently
viewed in terms of substitution, or when one
party receives the lion’s share of votes from a party
that no longer appears on the ballot. This is what happened
when the neo-fascist Italian Social Movement
(MSI) was absorbed into the National Alliance (AN)
on the right side of the Italian political spectrum in
the mid-1990s. Marginalized for most of the post-World
War II period, MSI support consisted of a broad mix of
social groups, including the under and unemployed,
civil servants and the urban- and petty-bourgeoisie, primarily
located in the south, and in particular, in urban
areas of the south (Ignazi, 1989; Tarchi, 1997).
As the Italian party crisis unfolded and left the MSI
unscathed, the leadership of the far-right sought to temper
and change the party’s image from a neo-fascist,
Mussolini-era holdover to a post-fascist, reformed and
European-style party. In part a strategy to advance at
294 M.E. Shin, J. Agnew / Electoral Studies 26 (2007) 287e302
the polls, especially since the disappearance of the
Christian Democrats created a vacuum in the center of
the Italian political spectrum, the MSI-AN was the
only viable conservative party on the 1994 ballot
throughout much of the center and south of Italy.
Aligned with Berlusconi’s center-right Forza Italia,
the AN has consistently outperformed its predecessor
in each election since 1992 (see Table 1). In this respect,
the AN not only inherited the majority of MSI support,
but it also appears to be mobilizing new voters and former
supporters of the disappeared DC and Italian socialist
party (Agnew, 1997).
Fig. 2 portrays how the AN replaced the MSI geographically.
The 1992 panel (upper left) serves as
a benchmark and maps significant local Moran’s I
values for the MSI the last time that it appeared on
the ballot.
Statistically significant regional clusters of high support
(solid black circles) are present around Rome and
in the Puglia region (the heel of Italy), while a large
cluster of low support (solid white circles) fans out
across northern Italy. In northwest Italy there are
a few spatial outliers, or provinces that contain high
support for the MSI, but are surrounded by provinces
with low MSI support (i.e., small black circles within
larger white circles). Prior to comparing the 1992 map
to the 1994, 1996 and 2001 maps, it is important to recall
that the bivariate form of the local Moran’s I index
is invoked for the next three panels in Fig. 2 and thus requires
a slightly different interpretation.
: Low values surrounded by high values
: Low values surrounded by low values
: High values surrounded by high values
: High values surrounded by low values
MSI
1992
MSI-AN
1994
AN
1996
AN
2001
Fig. 2. Substitution: MSI (1992), MSI-AN (1994) and AN (1996, 2001).
295M.E. Shin, J. Agnew / Electoral Studies 26 (2007) 287e302
Regional clusters of high support around Rome and
in Puglia (solid black circles) indicate that high levels
of support for the AN in 1994, 1996 and 2001 are surrounded
by similarly high levels of 1992 MSI support,
thus suggesting a degree of spatiotemporal stability between
the old MSI and new AN. The cluster of solid
white circles in northern Italy denotes low levels of
AN support in 1994, 1996 and 2001 are surrounded
by similarly low levels of 1992 MSI support, suggesting
that much of the north remains unreceptive to the postfascist
Italian right, despite its efforts to broaden its appeal.
There are relatively few spatial outliers, or provinces
with high AN support surrounded by low 1992
MSI support (i.e., small black circles within larger
white circles) or conversely, low AN support surrounded
by high 1992 MSI support (i.e., small white circles
surrounded by large black circles). Overall, the
continuity of clustering of MSI and AN support between
1992 and 2001, and the relative lack of spatial
outliers, illustrates how in this case replacement may
be analogous to substitution.
3.2. Replacement as splitting: from PCI to
PDS þ RC
As an extension of the previous example of substitution,
the replacement of the Italian Communist Party in
the early 1990s involved the splitting of votes, initially,
between the Democratic Party of the Left (PDS) and the
Communist Refoundation party (RC). The more moderate
and progressive PDS inherited the majority of infrastructure
and electoral base of the PCI, while the
smaller RC appealed to former-PCI militants and other
small, marginalized groups (Baccetti, 1997; Shin,
2001). One of the distinctive features of Italian communism
is the geography of its support, and in particular,
the bastion of leftist support in north-central Italy. In
fact, the administrative regions of Tuscany, Emilia-Romagna
and Umbria are collectively referred to as the
‘red belt’ or ‘red zone’. Fig. 3 reinforces this image,
and illustrates party replacement as the splitting of votes
between two or more parties. Again, the first panel (upper
left) in Fig. 3 serves as the benchmark, and maps
LISA statistics for the PCI the last time it appeared on
the ballot in 1987.
The red belt of the PCI is clearly visible as a patch of
solid black circles in the middle of the Italian peninsula,
where high levels of support for the PCI are significantly
clustered. The lone spatial outlier of low PCI
support surrounded by high PCI support (i.e., small
white circle within large black circle) in the red zone
is the province of Lucca. Due in part to legacies of small
landownership, and Lucca’s historical status as an independent
republic, the Italian left has always had difficulties
garnering support in this province (see Agnew,
1992; Dogan, 1967). A significant cluster of low PCI
support is visible in the northeast of Italy, and coincides
with an area referred to as the ‘white zone’ where the
Christian Democrats (DC) fared particularly well in
the post-World War II period.
The remaining bivariate LISA maps explore the relationship
between aggregated support for the PDS and
RC in 1992 and 1994, aggregated support for the DS,
RC and center-left Margherita or Daisy party in 2001,
and 1987 PCI support (results for 1996 are omitted
due to space limitations). The purpose of aggregating
support for parties of the new Italian left is to illustrate
party replacement as the splitting of old PCI votes. Note
that between 1996 and 2001, the PDS renamed itself the
Democrats of the Left (DS), the RC splintered into two
parties and that the Daisy party emerged as a promising
option for many voters on the center-left of the Italian
political spectrum. Notwithstanding such name
changes, party splits and new additions to the Italian
left, the map sequence in Fig. 3 is remarkably consistent.
High levels of support for the parties of the left,
namely the PDS and RC in 1992 and 1994, and the
DC, RC and Daisy in 2001 are surrounded by similarly
high 1987 PCI values in central Italy. The northeast remains
an area of low support for the Italian left in spatiotemporal
terms, though two significant spatial
outliers of high support surrounded by low support
emerge in 2001. The stability of clusters in Fig. 3 was
expected, but the two outliers in northeast Italy may indicate
that the new Italian left is making inroads into
what was once a DC stronghold.
3.3. Replacement as colonization/mobilization:
from DC to FI/LN
The previous examples of party replacement as substitution
and replacement as the splitting of votes are
relatively straightforward. A clear and definite lineage
between old and new party or parties can be identified
and is plausibly maintained in electoral terms in several
areas of Italy. But what happens when there is no direct
political descendant? This is what happened when the
Christian Democrats (DC) disintegrated in the early
1990s. As the leading party in government from 1948
to 1993, the DC enjoyed and took advantage of the
spoils of government until investigations publicized
the scope and extent of its corruption. The fallout
from the investigations precipitated the demise of the
party and its partner in crime, the Italian socialists,
296 M.E. Shin, J. Agnew / Electoral Studies 26 (2007) 287e302
and left a void in the middle of the Italian political spectrum
(see Gundle and Parker, 1996; Katz and Ignazi,
1996).
Into this void and materializing out of nowhere only
two months before the 1994 elections was Silvio Berlusconi’s
Forza Italia, which is considered by many to
be the primary replacement for the DC. Using the national
football chant as party name, Berlusconi effectively
used his privately owned television stations
during the short campaign, and shrewdly forged alliances
with the AN and regionalist Northern League.
More populist movement than political party, Forza Italiadrather
Berlusconidsuccessfully campaigned on
the promise of new jobs, prosperity, small government
and the continued threat of communism reincarnated
as the new Italian left. To determine whether or not
and how Forza Italia replaced the DC, LISA and bivariate
LISA maps are again utilized. The first panel (upper
left) in Fig. 4 shows clusters of DC support in 1992, the
last time that it appeared on the ballot.
Levels of support for the DC in Italy were clearly divided;
strong support was clustered in the south and
low support was concentrated in the center (red belt)
and northwest. Note that northeast Italy was once renowned
for its Catholic subculture, staunch DC support,
and was frequently seen as equivalent to the
communist subculture in central Italy (see Cartocci,
1990; Trigilia, 1986). The lack of significant DC clusters
in the northeast in 1992 may suggest that its historic
base became increasingly alienated from the
: Low values surrounded by high values
: Low values surrounded by low values
: High values surrounded by high values
: High values surrounded by low values
PCI
1987
PDS+RC
1992
PDS+RC
1994
DS+RC+MRG
2001
Fig. 3. Splitting: PCI (1987); PDS + RC (1992, 1994); DS + RC + MRG (2001).
297M.E. Shin, J. Agnew / Electoral Studies 26 (2007) 287e302
party some time before it finally disappeared. Examination
of the bivariate LISA maps between the DC and
Forza Italia reveals significant clusters of spatiotemporal
instability. In the northwest, where the DC performed
relatively poorly, support for Forza Italia is
relatively strong and significantly concentrated. Conversely,
in the south, where high levels of DC support
clustered in 1992, support for Forza Italia is below average,
especially in 1994. The only region where high
support for Forza Italia overlaps historic support for
the DC is in Sicily. Though Forza Italia may have replaced
the DC in some areas of Italy, replacement
was not wholesale as indicated by the clusters of spatiotemporal
instability.
3.4. Estimating party replacement in the Veneto,
a worked example
The exploratory spatial data analyses illustrate how
the process of party replacement varies geographically,
and challenges the utility of national-level flow-of-vote
or voter transition estimates. To assess the plausibility
of the arguments made here regarding how the process
of party replacement occurred in Italy, voter transitions
from the DC to Forza Italia and the Northern League in
the Veneto are estimated using Gary King’s (King,
1997, 2002) method of ecological inference. Other
methods, such as entropy maximization (see Johnston
and Pattie, 2000), can be used to estimate such
: Low values surrounded by high values
: Low values surrounded by low values
: High values surrounded by high values
: High values surrounded by low values
DC
1992
FI
1994
FI
1996
FI
2001
Fig. 4. Colonization/Mobilization: DC (1992); Forza Italia (1994, 1996, 2001).
298 M.E. Shin, J. Agnew / Electoral Studies 26 (2007) 287e302
transitions, but King’s method has a relatively low
threshold in terms of data requirements and was used
previously to examine national-level voter transitions
in consecutive elections in Italy (see Benoit et al.,
2004; Wellhofer, 2001), but not party replacement per
se. Based upon constrained maximum likelihood analysis
(see King, 1997), party replacement estimates are
obtained from the identity Ti ¼ bb
i Xi þ bw
i ð1 À XÞ
where the independent variable, X, is the vote share
for the old party in the first or baseline election, and
the dependent variable, T, is the vote share for the new
party in a subsequent election. The quantity of particular
interest is bb
i , the percentage of voters who supported the
old party in the baseline election and who also voted for
the new party in a subsequent election.
Located in northeastern Italy, the Veneto was one of
the strongholds for the DC, and is one of the regions
where the regionalist Northern League finds its greatest
support. Though no significant amounts of spatiotemporal
clustering are visible in the northeast of Italy in
Fig. 4, initial levels of League support more or less mirrored
that of the former DC, and it is often presumed
that the League is substituting for the DC in the Veneto
and across northern Italy (e.g., Diamanti, 2003; Golden,
2004). Though divisions within the party and its base
have recently marginalized what once was the largest
party in northern Italy, its alliance with Berlusconi’s
Forza Italia gives the League continued political relevance.
Using election data from the 582 municipalities
of the Veneto, Table 3 reports provincial estimates of
voter transitions from the DC in 1992 to the League
and Forza Italia in 1994, 1996 and 2001, or in other
words, the proportion of voters who supported the DC
in 1992 and who also voted for the League or FI in
1994, 1996 or 2001. Note that 1992 is used as the baseline
year for comparison because it is the last year the
DC appeared on the ballot and we are specifically interested
in how the DC was replaced.
Several interesting features emerge from the table of
voter transitions. First, the regional transition estimates
mask notable provincial (i.e., sub-regional) differences.
For example, at the regional level it is estimated that
28% of former-DC voters voted for the League in
1994. In the province of Rovigo, however, the estimate
is only 13%, but in Venice and Verona the estimate approaches
40%. Second, there are notable variations in
the estimates over time. In most provinces, both the
League and FI receive moderate levels of support
from former-DC voters in 1994, in 1996 it appears
that former-DC voters tended to favor the League, but
in 2001 there was a reversal of electoral fortunes with
former-DC voters supporting Forza Italia over the
League. Closer inspection reveals some notable transitions,
such as that from the DC to Forza Italia in Verona.
In 1994 and 1996, it is estimated that only 5% of
DC supporters voted for Berlusconi’s party, but in
2001 this figure is 26%. Such a dramatic increase is consistent
with the argument that Forza Italia is colonizing
former-DC space and effectively mobilizing voters. The
geographical instability and temporal volatility of voter
transition estimates in Table 3 illustrates clearly that the
process of party replacement and electoral change
across the Veneto is far more complex than the simple
and direct substitution of the DC with the Northern
League, and more generally, that a geographic framework
can provide useful insights into the process of
electoral change.
4. Discussion and conclusion
Party replacement is not equivalent to substitution
and arguably involves an array of diverse processes.
Moreover, the electoral processes underlying party replacement
are geographic in nature. For instance, it is
often presumed that the Northern League replaced the
Christian Democrats throughout northern Italy, and in
some places this is indeed what occurred. However,
the spatiotemporal instability of party replacement estimates
suggests that while the League was able to attract
former-DC voters and to colonize some provinces of the
Veneto, it was also a poor choice among other possible
DC replacements, such as Forza Italia.
Table 3
Party replacement estimates for the DC to LN/FI in the Veneto, 1992e
2001
Voter transitions from the DC, 1992
LN94 LN96 LN01 FI94 FI96 FI01
Belluno 0.37 0.57 0.12 0.18 0.28 0.31
0.06 0.07 0.25 0.04 0.04 0.04
Padova 0.33 0.52 0.18 0.29 0.19 0.47
0.02 0.03 0.00 0.02 0.02 0.03
Rovigo 0.13 0.32 0.11 0.24 0.22 0.35
0.01 0.02 0.01 0.00 0.04 0.05
Treviso 0.32 0.57 0.21 0.28 0.18 0.51
0.02 0.04 0.02 0.01 0.00 0.02
Venezia 0.37 0.66 0.19 0.28 0.14 0.46
0.01 0.04 0.01 0.05 0.04 0.06
Verona 0.38 0.64 0.27 0.05 0.05 0.26
0.04 0.02 0.01 0.02 0.01 0.03
Vicenza 0.22 0.47 0.30 0.25 0.14 0.51
0.04 0.04 0.01 0.02 0.02 0.04
Veneto 0.28 0.41 0.21 0.22 0.11 0.47
0.02 0.02 0.01 0.01 0.01 0.01
Standard errors reported below each estimate.
299M.E. Shin, J. Agnew / Electoral Studies 26 (2007) 287e302
Explanations for the popularity of the League, and in
particular, its ability to colonize new ‘spaces’ in the
Veneto are in part linked to increasing levels of secularization
in the region and the failure of the DC to support
the small firms and enterprises that are credited with
bolstering the Italian economy over the last quarter century
(Riccamboni, 1997, 299). The stagnating and corrupt
DC model of representation, which relied upon
small towns and rural agricultural areas, failed to recognize
and address the needs of the growing northern Italian
economy that was characterized by high levels of
employment, rising productivity and industrialization
(Diamanti, 1993). It is precisely in the industrial districts
and the ‘urbanized countryside’ of the north where
the League’s platform based upon ‘localismo’, or the
preservation, promotion and protection of local territorial
identities and culture, resonated the most with
voters.
The example of the League illustrates clearly the
need to consider the geographical bases of party replacement
and political change, but similar arguments
can be made when replacement is akin to substitution.
Referring back to the replacement of the neo-fascist
Italian Social Movement (MSI) with the post-fascist
National Alliance (AN), spatial analyses reveal that
the geographies of support for each party more or less
overlap. Notwithstanding this overlap, there were also
significant regional differences in support: the Italian
right is popular throughout much of the south, but its
electoral performance in the north is mediocre at best.
The detection of such clustering is important when examining
and explaining political change because it is
often presumed that political change is a stationary process.
For instance, when using flow of vote estimates to
evaluate political changes at the national level, it is implied
that vote shifts between parties are consistent in
magnitude and direction across all units of analysis,
and consequently, that units of analysis are more or
less identical and interchangeable with one another
(Agnew, 1996). This assumption is often untenable
and in some cases it may be problematic because electoral
shifts are likely to be under- or over-estimated, depending
upon where they are obtained. In other words,
political change is seldom uniform across a democracy.
As illustrated in the worked example, using a spatial analytic
framework prior to obtaining estimates of vote
flows and voter transitions may improve estimates and
yield a more accurate account of electoral change.
Italian democracy presents itself as a compelling
arena to examine political change because the recent
transformation of parties was so dramatic and comprehensive.
Spatial analyses show that the nature of
political change in Italy was, and continues to be, geographically
dependent. The variety of clusters detected
indicates that different processes govern change in different
places across Italy. The spatial analytic approach
used in this study is not limited to the Italian case or to
party replacement, but can be applied elsewhere to examine
electoral change in general. Showing how electoral
change occurs, and being able to locate precisely
such changes, gives a more complete understanding of
how political change occurs.
Acknowledgements
This research was supported by a grant from the National
Science Foundation (NSF-BCS 0322334).
Thanks are also extended to the Cattaneo Institute in
Bologna, Italy, for providing the election data and to
the anonymous reviewers for their insights on a previous
version of the paper.
References
Agnew, J., 1987a. Place and Politics: The Geographical Mediation of
State and Society. Allen & Unwin, London.
Agnew, J., 1987b. Place anyone? A comment on the McAllister
and Johnston papers. Political Geography Quarterly 6 (1),
39e40.
Agnew, J., 1988. ‘Better thieves than Reds’? The nationalization thesis
and the possibility of a geography of Italian politics. Political
Geography Quarterly 7, 307e321.
Agnew, J., 1992. Place and politics in post-war Italy: a cultural geography
of local identity in the provinces of Lucca and Pistoia. In:
Anderson, K., Gale, F. (Eds.), Inventing Places: Studies in Cultural
Geography. Longman Cheshire, Melbourne, pp. 52e71.
Agnew, J., 1995. The rhetoric of regionalism: the Northern League in
Italian Politics, 1983e1994. Transactions of the Institute of British
Geographers 20, 156e172.
Agnew, J., 1996. Mapping politics: how context counts in electoral
geography. Political Geography 15, 129e146.
Agnew, J., 1997. The dramaturgy of horizons: geographical scale and
the ‘Reconstruction of Italy’ by the new Italian political parties,
1992e1995. Political Geography 16 (2), 99e121.
Agnew, J., 2002. Place and Politics in Modern Italy. University of
Chicago Press, Chicago.
Agnew, J., Shin, M.E., Bettoni, G., 2002. City versus metropolis: the
Northern League in the Milan metropolitan area. International
Journal of Urban and Regional Research 26 (2), 266e283.
Anselin, L., 1988. Spatial Econometrics: Methods and Models.
Kluwer Academic Publishers, Dordrecht.
Anselin, L., 1989. What is special about spatial data? Alternative perspectives
on spatial data analysis. Symposium on Spatial Statistics,
Past, Present and Future. Department of Geography,
Syracuse, NY.
Anselin, L., 1992. Spatial data analysis with GIS: an introduction to
application in the social sciences. National Center for Geographic
Information Analysis (NCGIA), Santa Barbara, CA.
Anselin, L., 1995a. Local indicators of spatial association e LISA.
Geographical Analysis 27, 93e115.
300 M.E. Shin, J. Agnew / Electoral Studies 26 (2007) 287e302
Anselin, L., 1995b. SpaceStat, Version 1.80 User’s Guide. Regional
Research Institute, Morgantown, WV.
Baccetti, C., 1997. Il Pds. Il Mulino, Bologna.
Bartolini, S., D’Alimonte, R. (Eds.), 1995. Maggioritario ma non
Troppo. Il Mulino, Bologna.
Benoit, K., Laver, M., Giannetti, D., 2004. Multiparty split-ticket
voting estimation as an ecological inference problem. In:
King, G., Rosen, O., Tanner, M. (Eds.), Ecological Inference:
New Methodological Strategies. Cambridge University Press,
Cambridge, pp. 333e350.
Burnham, W.D., 1970. Critical Elections and the Mainsprings of
American Politics. Norton, New York.
Butler, D., Stokes, D., 1969. Political Change in Britain. St. Martin’s,
New York.
Caciagli, M., Spreafico, A., 1990. Vent’anni di Elezioni in Italia.
1968e1987. Liviana, Padova.
Calvo, E., Escolar, M., 2003. The local voter: a geographically
weighted approach to ecological inference. American Journal
of Political Science 47 (1), 189e204.
Cartocci, R., 1990. Elettori in Italia. Il Mulino, Bologna.
Cartocci, R., 1994. Fra Lega e Chiesa. Il Mulino, Bologna.
Cho, W.T., 2003. Contagion Effects and Ethnic Contribution Networks.
American Journal of Political Science 47, 368e387.
Cliff, A., Ord, J., 1981. Spatial Processes: Models and Applications.
Pion Limited, London.
D’Alimonte, R., Bartolini, S. (Eds.), 2002. Maggioritario Finalmente?
La Transizione Elettorale 1994e2001. Il Mulino, Bologna.
D’Alimonte, R., Chiaramonte, A., 1995. Il nuovo sistema elettorale
Italiano: le opportunita` e le scelte. In: Bartolini, S.,
D’Alimonte, R. (Eds.), Maggioritario ma non Troppo. Il Mulino,
Bologna, pp. 37e81.
Diamanti, I., 1993. La Lega. Geografia, Storia e Sociologia di un
Soggetto Politico. Donzelli Editrice, Roma.
Diamanti, I., 1996. Il Male del Nord: Lega, Localismo, Secessione.
Donzelli editore, Roma.
Diamanti, I., 2003. Bianco, Rosso, Verde . e Azzurro. Mappe e Colori
dell’Italia Politica. Il Mulino, Bologna.
Diamanti, I., Mannheimer, R., (Eds.), 1994. Milano a Roma: Guida
all’Italia elettorale del 1994. Donzelli Editore, Roma.
Dogan, M., 1967. Political cleavages and social stratification in
France and Italy. In: Lipset, S., Rokkan, S. (Eds.), Party Systems
and Voter Alignments. The Free Press, New York, pp. 129e195.
Foot, J., 1996. The ‘Left Opposition’ and the Crisis: Rifondazione
Comunista and La Rete. In: Gundle, S., Parker, S. (Eds.), The
New Italian Republic. Routledge, London, pp. 173e190.
Galli, G., Prandi, A., 1970. Patterns of Political Participation in Italy.
Yale University Press, New Haven.
Gangemi, G., Riccamboni, G. (Eds.), 1997. Le Elezioni della Transizione.
UTET Libreria, Torino.
Golden, M., 2004. International economic sources of regime change:
how European integration undermined Italy’s postwar party system.
Comparative Political Studies 37, 1238e1274.
Griffith, D., 1987. Spatial Autocorrelation: A Primer. Association of
American Geographers, Washington, DC.
Gschwend, T., Johnston, R.J., Pattie, C., 2003. Split-ticket patterns in
mixed-member proportional elections systems: estimates and
analyses of their spatial variation at the German federal election,
1998. British Journal of Political Science 33, 109e127.
Gundle, S., Parker, S., (Eds.), 1996. The New Italian Republic: From
the Fall of the Berlin Wall to the Berlusconi. Routledge, London.
Haining, R., 2003. Spatial Data Analysis: Theory and Practice. Cambridge
University Press, Cambridge.
Hellman, S., 1992. The difficult birth of the Democratic Party of the
Left. In: Hellman, S., Pasquino, G. (Eds.), Italian Politics: A Review,
vol. 7. Pinter, London, pp. 68e86.
Ignazi, P., 1989. Il Polo Escluso: Profilo Storico del Movimento Sociale
Italiano. Il Mulino, Bologna.
Ignazi, P., 1992. Dal Pci al Pds. Il Mulino, Bologna.
Janda, K., 1980. Political Parties: A Cross-national Survey. The Free
Press, New York.
Johnston, R.J., 1986. The neighborhood effect revisited: spatial science
or political regionalism? Environment and Planning D 4,
41e55.
Johnston, R.J., 1992. A Question of Place. Blackwell, Oxford.
Johnston, R.J., Pattie, C., 1992. Class dealignment and the regional
polarization of voting patterns in Great Britain, 1964e1987. Political
Geography 11, 73e86.
Johnston, R.J., Pattie, C., 1999. Constituency campaign intensity and
split-ticket voting: New Zealand’s first election under MMP,
1996. Political Science 51 (2), 164e181.
Johnston, R.J., Pattie, C., 2000. Ecological inference and entropy
maximizing: an alternative procedure for split-ticket voting. Political
Analysis 8, 333e345.
Katz, R., 1973. The attribution of variance in electoral returns: an alternative
measurement technique. American Political Science Review
67, 817e828.
Katz, R., Ignazi, P. (Eds.), 1996. Italian Politics: The Year of the Tycoon.
Westview Press, Boulder.
Katz, R., Mair, P. (Eds.), 1994. How Parties Organize: Change and
Adaptation in Party Organizations in Western Democracies.
Sage, London.
Kertzer, D., 1996. Politics and Symbols: the Italian Communist Party
and the Fall of Communism. Yale University Press, New Haven.
Key, V.O., 1955. A theory of critical elections. The Journal of Politics
17 (1), 3e18.
Kim, J., Elliot, E., Wang, D.-M., 2003. A spatial analysis of countylevel
outcomes in US Presidential elections: 1988e2000. Electoral
Studies 22, 741e761.
King, G., 1996. Why context should not count. Political Geography
15, 159e164.
King, G., 1997. A Solution to the Ecological Inference Problem.
Princeton University Press, Princeton.
King, G., 2002. EI: A Program for Ecological Inference v. 2.65. Harvard
University, Cambridge.
Kitschelt, H., 1992. The formation of party systems in East Central
Europe. Politics & Society 20 (1), 7e50.
Kitschelt, H., 1995. Formation of party cleavages in post-communist
democracies. Party Politics 1 (4), 143e160.
McAllister, I., 1987. Social context, turnout and the vote: Australian
and British comparisons. Political Geography Quarterly 6 (1),
17e30.
McCarthy, P., 1996. Forza Italia: the overwhelming success and consequent
problems of a virtual party. In: Katz, R., Ignazi, P. (Eds.),
Italian Politics: The Year of the Tycoon. Westview Press, Boulder.
McCarthy, P., 1997. The Crisis of the Italian State. Macmillan Press,
London.
Morlino, L., 1996. Crisis of parties and change of party system in
Italy. Party Politics 2 (1), 5e30.
Nardulli, P., 1995. The concept of a critical realignment, electoral behavior,
and political change. American Political Science Review
89 (1), 10e22.
Nelken, D., 1996. A legal revolution? The judges and Tangentopoli.
In: Gundle, S., Parker, S. (Eds.), The New Italian Republic. Routledge,
London.
301M.E. Shin, J. Agnew / Electoral Studies 26 (2007) 287e302
O’Loughlin, J.V., Flint, C., Anselin, L., 1994. The geography of the
Nazi vote: context, confession and class in the Reichstag election
of 1930. Annals of the Association of American Geographers 84,
351e380.
O’Loughlin,J.V.,Shin,M.E.,Talbot, P.,1996.Politicalgeographiesand
political cleavages in the Russian parliamentary elections of 1993
and 1995. Post-Soviet Geography and Economics 37, 355e386.
O’Sullivan, D., Unwin, D., 2003. Geographic Information Analysis.
John Wiley & Sons, Hoboken.
Panebianco, A., 1988. Political parties: organization and power.
Cambridge University Press, Cambridge.
Pappalardo, A., 1995. La nuova legge elettorale in parlamento: chi,
come e perche`. In: Bartolini, S., D’Alimonte, R. (Eds.), Maggioritario
ma non Troppo. Il Mulino, Bologna, pp. 13e36.
Pattie, C., Johnston, R.J., 1997. Local economic contexts and changing
party allegiances at the 1992 British General Election. Party
Politics 3 (1), 79e96.
Pattie, C.J., Johnston, R.J., 2003. Local battles in a national landslide:
constituency campaigning at the 2001 British General Election.
Political Geography 22 (4), 381e414.
Riccamboni, G., 1997. Ritorno al futuro? La transizione nell’ex subcultura
‘‘bianca’’. In: Gangemi, G., Riccamboni, G. (Eds.), Le
elezioni della transizione. Il sistema politico italiano alla prova
del voto 1994e1996. UTET, Torino, pp. 279e301.
Shin, M.E., 2001. The politicization of place in Italy. Political Geography
20, 331e352.
Shin, M.E., Agnew, J., 2002. The geography of party replacement in
Italy, 1987e1996. Political Geography 21, 221e242.
Tarchi, M., 1997. Dal Msi ad An. Il Mulino, Bologna.
Trigilia, C., 1986. Grandi Partiti e Piccole Imprese. Il Mulino,
Bologna.
Wellhofer, E.S., 2001. Party realignment and voter transition in
Italy, 1987e1996. Comparative Political Studies 34 (2),
156e186.
302 M.E. Shin, J. Agnew / Electoral Studies 26 (2007) 287e302