Multi-level political change: Assessing electoral volatility in 58 European regions (1993-2022)

In recent years, increasing attention has been devoted to electoral turmoil associated with the emergence of new political actors or the crisis of established parties at national level. However, in multi-level systems, transformations have also affected sub-national politics. This article seeks to understand some hitherto unexplored aspects of political change in regional party systems. Change is linked to the concept of instability and operationalized in terms of electoral volatility. To account for the fact that instability may be driven by distinct regional and national pressures coexisting in regional elections, a new measure of volatility has been developed. It is disaggregated into two territorial components: ‘Region-specific volatility’ (RSV) and ‘Region-transcending volatility’ (RTV). RSV refers to changes in electoral support for political parties competing exclusively in one region, while RTV is calculated for parties that ‘transcend’ regional boundaries – i.e. they are electorally active in several/all regions of a country or are part of institutionalized inter-regional networks. By applying this new measure to 385 elections in 58 European regions, this article shows that levels and types of instability have varied significantly over time, across regions and across countries. This may also account for different developments in territorial politics observed in Western Europe.

Making an analytical distinction between RSV and RTV allows us to better understand how political changeresulting from electoral destabilizationoccurs in multi-level systems.While the instability of national party systems plays a crucial role in shaping state-wide democratic processes and outcomes, regions may follow different paths: even when they experience profound political change, this may not be as rooted in what happens at the national level.
The next section explores the concept of instability by focusing on volatility in the electoral arena.It adapts frameworks developed by scholars interested in national politics to the study of regional contexts.The two components of regional volatility -RSV and RTVare defined and measured.They are then applied to almost 400 elections held in 58 regions of four countries -Italy, Spain, Germany and the UKbetween 1993 and 2022.The choice of a 30year time span also facilitates comparisons between precrisis (1993-2007) and post-crisis (2008-2022) periods of similar length.The key aim of the article is to 1) develop a new measure of volatility accounting for its territorial components and 2) show that, on the whole, regional political systems have indeed become more unstable but 3) there is considerable variation in the extent to which this instability depends on vote shifts between regiontranscending or region-specific parties.The latter result can also be used to illustrate that, more generally, the territorial politics of the four major Western European countries have followed distinct trajectories.Recognizing the different territorial roots of instability is therefore a prerequisite for future studies that seek to investigate the true causes and consequences of multi-level political change.

Instability in national and regional elections
Ever since the seminal article published by Mogens N. Pedersen in 1979, volatility has been commonly used as a measure of instability and change from one election to the next in national party systems.Already in the early 1990s, Bartolini and Mair (1990) presented a comprehensive assessment of voters' 'electoral availability' in European democracies.Although they noted limited episodes of significant turbulence, they concluded that Western European electorates remained generally stable at the end of the 20 th century.Importantly, the two scholars made a distinction between 'within-block' and 'between-block' volatility, highlighting that a significant part of the vote shifts observed in European democratic systems occurred among parties belonging to the same family or positioned on the same side of a political 'cleavage'.Measures of volatility have then been used extensively, adapted and refined by various scholars, also to assess the stability of new democracies in different parts of the world (Casal Bértoa et al., 2017).Keeping the focus on Western Europe, Chiaramonte andEmanuele (2017, 2022) have more recently relied on volatility as one of the key indicators to measure broader processes of de-institutionalization of party systems. 1 Reviewing the literature and considering the different methods used to assess political change, they emphasized another important analytical distinction: that between alteration volatilityvote switching across existing partiesand regeneration volatilitycaused by the entry of new parties and the exit of old ones from the party system.
In general, it is not uncommon for research on electoral change to define new analytical dimensions that disaggregate total volatility into more specific categories in order to identify the real sources of instability.For Bartolini and Mair, only inter-block volatility could indicate fundamental transformations in the cleavage structure of democratic systems.For Chiaramonte and Emanuele, distinguishing between alteration and regeneration volatility is key in determining whether a deeper process of deinstitutionalization of party politics is occurring in Western Europe.So far, however, no territorial specification of electoral instability has been conceptualized, operationalized and systematically applied to a wide range of cases.This is surprising given that party competition in most democratic systems occurs at different levels of government: it is indeed multi-level.
While studies on electoral (in-)stability have increased in number and scope in recent years, another (parallel) stream of research on sub-national party politics has also flourished.Interest in local and regional elections has contributed to the development of theories focusing on the vertical linkages and electoral/representational gaps between national and subnational arenas of party competition (Schakel, 2011;Schakel and Romanova, 2020;Thorlakson, 2020).Generally, studies on regional politics have paid significant attention to 'dissimilar' or 'congruent' regional and national political outcomes.Dissimilarity has been mainly explained in structural terms by referring to the depth of relatively stable territorial cleavages underpinning the mobilization of regionalist and non-state-wide parties.Some authors have also tried to explain over-time electoral change in regional elections but have mainly concentrated on short-term swings and the presence of 'second-order' electoral effects (Schakel, 2011(Schakel, , 2021;;Schakel and Jeffery, 2013).So far no study has systematically analysed longer-term change in regional politics, occurring independently of electoral-cycles and institutional asymmetries.One question that has not yet been directly addressed is whether regional party systems are becoming inherently more unstable.Moreover, it is still unclear whether sub-national instability is just a reflection of the increasing fluidity affecting state-wide parties or also has more regionspecific roots.
To fill the gaps highlighted above, this article suggests adding a new analytical dimension to the study of volatility.Bolgherini et al. (2021) have already pointed out that electoral changes from national to regional elections can be Vampa nationally or regionally driven, depending on whether they are caused by shifts in support for parties competing in the national arena or result from the success/collapse of regionspecific parties.Although in the article by Bolgherini et al. this differentiation is mainly used to assess the 'secondorderness' of Italian regional elections in the 2018-2021 cyclefollowing the example of other scholars (Dandoy and Schakel, 2013;Schakel, 2011) their framework can be further developed to create a new measure of instability from one regional election to the next and apply it to a longer period of time.Focusing on the regional level, an analytical distinction can be made between 'region-specific volatility' (RSV) and 'regiontranscending volatility' (RTV).This distinction recognizes the fact that regions are 'intermediate' arenas of political competition in which stability and change may derive from both state-wide and region-specific dynamics.
To quantify RSVand RTV correctly and consistently, it is necessary to establish clear criteria for the selection of parties to be considered for either type of volatility.A party is classified as 'region-specific' (and thus included in the RSV calculations) if it (1) only competes in one region, (2) explicitly refers to the region as its main arena of political action and (3) does not act as a 'functional equivalent' of a state-wide party or similar parties active in other regions.A party can compete in both regional and general elections and still be region-specific if its electoral base is concentrated in one region.
A region can be institutionally or historically/culturally defined (Schmitt-Egner, 2002).In most cases, the two aspects coincide, but sometimes they do not, and this can also play an important role in the mobilization of region-specific parties.For instance, some regionalist parties based in the Autonomous Community of the Basque Country in Spain are also activewith mixed resultsin neighbouring Navarre, which they regard as part of the greater Basque region (a set of territories with shared history and cultural traditions).Despite transcending the institutionally defined boundaries of the 'core' Basque Autonomous Community, it is clear that Basque regionalist parties still focus on a geographically and historically defined region (Schmitt-Egner, 2002: 188).Therefore they can be categorized as 'region-specific'.
The case of the Northern League (later renamed 'League') in Italy is quite different.Since its foundation, the party has been continuously active in all regions of Northern and Central Italy and has even been present in the Southalso being included in national governments for a significant number of years (Albertazzi and Vampa, 2021).Rather than a regionalist party, the League has been defined as a 'macroregionalist' party (McDonnell and Vampa, 2016).Given its origins and evolutiontoday the party is a full-fledged state-wide party (Albertazzi et al., 2018) the League can thus be categorized as a 'region-transcending party'.
The concept of 'functional equivalence' (Van Deth, 1998) can be applied to parties that, despite operating in just one region and having strong links with a regional community, formally act as representatives of and forge stable alliances with other parties with similar ideological orientations and active in other regions.This happens, for example, when the Christian Social Union (CSU) competes in the Bavarian elections in place of the German Christian Democratic Union (CDU), with which it then forms a single group in the federal parliament in Berlin.Similarly, in Spain the Union of the Navarrese People (UPN) has dominated the politics of Navarre (Vampa, 2020) while de facto acting as the regional branch of the Spanish People's Party (PP) or forging coalitions with it for most of its history. 2 In short, both CSU and UPN have been consistently integrated into national political structures and can therefore be classified as 'region-transcending'.
Borderline cases can be found in Northern Irelandthe Social Democratic Labour Party (SDLP) and Allianceand in Scotlandthe Scottish Greens.Some studies have defined SDLP and Alliance as 'sister parties' of the British Labour Party and Liberal Democrats (Henderson and Wyn Jones, 2021: 172).Yet SDLP is committed to Irish reunification and Alliance has been neutral on the issueand stayed in opposition in Westminster when the Lib Dems joined the coalition with the Conservative Party in 2010-2015.The Scottish Greens are completely separate from the Greens of England and Wales and, what is more, support Scottish independence.Thus, adopting a strict interpretation of the concept of functional equivalence, these three cases can be classified as region-specific, as they do not present a level of ideological-organizational integration with their British counterparts similar to that observable, for example, in the relationship between the CSU and CDUwhere one party clearly acts as agent and representative of another.
Having drawn the line dividing region-specific and region-transcending political camps, we are now able to assess more accurately the different manifestations of subnational electoral instability.By combining the two types of volatility, we end up with four different scenarios, as shown in Table 1 (here the framework adopted by Chiaramonte and Emanuele (2017) using volatility combinations to identify different stability/instability scenarios is particularly useful).We can start from a situation where small changes are observable in both camps (A) and therefore we have a stable regional party system.If instead we observe high volatility among region-transcending parties but not among regionspecific ones, then this suggests that instability is driven by competition between national (or macro-regional/functionally equivalent) parties (scenario B: nationally-driven instability).We have the opposite scenario, when electoral shifts mainly affect region-specific parties (C, regionallydriven instability).Finally, we could have a situation in which volatility is high both between region-specific parties and between region-transcending parties, in which case we can speak of 'multi-level' instability (D).
The next section applies the framework presented above to the regions of Italy, Spain, Germany and the United Kingdom.The main objectives of this exploratory exercise are: (1) to assess trends in regional electoral volatility; (2) to identify its territorial roots and the resulting stability/ instability scenarios; (3) to develop a multi-level analysis comparing patterns between and within countries.
Assessing to what extent regional instability is driven by region-specific or region-transcending political actors and how the two types of volatility are linked to each other is not only a valuable empirical exercise, but it also has important implications for the study of overlooked developments in contemporary party politics.Indeed, while growing regiontranscending volatility may keep regions aligned with simultaneous changes in national politics, rising regionspecific volatility may widen vertical electoral gaps, which cannot be adequately explained by theories focusing on institutional structures, well-established cleavages, mobilization of regionalist parties and election cycles.Additionally, synergies between the two types of volatility may in turn lead to an acceleration of the deinstitutionalization processes already observed nationally.Lastly, the policy effects of sub-national instability have not been sufficiently appreciated: what type of regional political change is more likely to alter national-regional relations and sub-national policies?While this paper does not provide answers to all these questions, it prepares the ground for future studies seeking to unpack regional political change and understand its real causes and effects.

Methodological note
The empirical analysis presented below is based on an original dataset that includes 385 regional elections from 1993 to 2022 in 58 regions in four countries -Italy, Spain, Germany and the UK. 3 These are countries of comparable size, where regional governments have enjoyed significant powers and regional politics also plays an important role in shaping party competition and electoral outcomes.This, of course, does not preclude a more general application of the framework presented here to other multilevel systems, as also emphasised in the conclusion.The choice to adopt a 30-year period is justified by the fact that it can in turn be divided into two sub-periods of equal length (15 years): one runs from 1993 to 2007, the other from 2008 to 2022.In this way, it will be possible to assess whether in the years since the financial crisiswhich was followed by other crises of various kindsregional politics has become more unstable than in the previous period and how this instability has manifested itself differently across regions and countries.
For each election the Pedersen index is used to calculate regional volatility and then its 'region-specific' (RSV) and 'region-transcending' (RTV) components.This is the general Pedersen index formula

Regional Volatility
where n is the total number of parties considered within two consecutive regional elections (at time t and t + 1) and p i represents the percentage of votes received by each parties at time t and t + 1.The resulting score ranges from 0, where no change is detected, to 100 if the whole political landscape changes from one regional election to the next.
Based on the definitions provided in the previous section, the Pedersen index can be more narrowly applied to parties competing only in one region to calculate RSV.If instead it is applied to parties that are electorally active in more regions or nationally, we obtain RTV.Therefore, regional volatility can also be expressed as the combination of the two While the Pedersen index is a useful tool for assessing changes in voting behaviour, it has some limitations.Specifically, it provides an aggregate measure that does not capture individual-level changes.Ideally, a study of volatility would incorporate an individual perspective.However, due to the lack of individual-level data across a large number of regions over an extended period of time, this study must rely on aggregate data.This approach, similar to that adopted by other scholars in the field (some quoted in the previous section), allows us to identify trends and important changes in the structural features of party systems, despite its imperfections.The following discussion of data and results is divided into three parts following a multi-level approach.First, the article provides a general overview of how stable/unstable regional elections have been from 1993 to 2022.Next, it examines the different patterns of regional electoral volatility between countries and within each country by relying on the distinction between RSV and RTV.Finally, it turns to a more comparative assessment, linking the diverse manifestations of regional instability to the different evolution of territorial politics in the four countries.
Results: electoral volatility in 58 regions

General trends
We first consider general trends in volatility in 58 regions in four countries.Figure 1 covers exactly three decades and is based on the results of 385 regional elections.Since regions do not all vote at the same time, the elections have been grouped into 5-year cycles.We can see that the 1990s started with high total volatility (dashed grey line).This was immediately followed by a sharp decline and stabilization.In the 2008-2012 period we see a new increase in volatility which reaches a peak in 2013-2017 but then, unlike the 1990s, does not decline and stabilize but remains high in the subsequent 2018-2022 cycle.RTV (solid grey line) which captures 'region-transcending volatility' between national (or macro-regional) parties is initially very close to total volatility but then it starts diverging from it.This is because an increasingly larger share of the rise in total volatility observed since 2008 is determined by RSV, that is, volatility affecting region-specific parties.The latter (black line) remains lower than RTV but follows a steady upward trend from 1993-1997 to 2018-2022.If instead of looking at the trends in election cycles, we divide the three decades into two periods of the same duration, 1993-2007 (194 regional elections) and 2008-2022 (191 regional elections), total volatility increases on average from 19.8% in the first period to 26.3% in the second.So it has increased substantially, by one third.RTV has increased by almost four percentage points, from 16.1 to 19.8%, while RSV, which was quite low in the 1990s, has almost doubled from 3.6 to 6.7%.All increases are statistically significant at any conventional level.Thus, compared to the 1990s and early 2000s, the post-financial crisis electoral environment of the regions included in the dataset has generally been characterized by a more sustained volatility, which does not seem to have resulted (yet) in a new equilibrium.
These first results may already lead us to the hasty conclusion that regional volatility in the major European decentralized countries has increased significantly in both its more region-specific and region-transcending componentsmeaning that both regional and national actors have driven sub-national political change.Yet the picture may appear more complex if we consider the different combinations between RSV and RTV.We can sort all 385 regional elections of our sample into the four scenarios identified in Table 1.The results are shown in Table 2 and compare distributions in 1993-2007 and 2008-2022.The cut-off points are represented by the two mean values of RSV (5) and RTV (17.9) across all regional elections of the four countries for the whole 1993-2022 period.We can see that while more than half of the regional elections held in 1993-2007 can be classified as territorially stable (quadrant A), only a third of the 2008-2022 elections occupy the same position.Instead, after 2007 we see a dramatic increase in elections characterized by multi-level instability (from 7.2 to 24.1% of regional elections), a slight increase in nationally-driven instability (from 24.3 to 25.7%) and a small decline in region-specific instability (from 18 to 15.7%).
These trends give us an overview of how political instability, operationalized in terms of electoral volatility, has evolved in 58 regions of four major European countries.However, behind this general picture may lie interesting variations that should be analysed from a truly multi-level perspective.In this way, it would also be possible to explore the full potential of the territorial measures of volatility developed in this article, without which we would not be able to identify and measure differences between countries and between regions within each country.

Cross-country variations
Starting from an analysis of cross-country variations Figures 2 and 3 show trends in RTV and RSV in Italy, Spain, Germany and Great Britain separately.We can see that most of the high volatility observed in the early 1990s in Figure 1 occurred in Italian regions, where RTV was very high (Figure 2) due to the collapse of the national political system at the end of the Cold War, after a series of corruption scandals (Morlino, 1996).Despite falling considerably in the late 1990s, RTV in the Italian regions remained relatively high and increased again after 2008.The 1990s and early 2000s were instead a period of relative stability in Spanish regions, where RTV started to increase significantly only in the post-2008 period, converging with Italy.In Germany, too, a trend of RTV growth can be observed, but at a more gradual pace since 1993-1997.Finally, RTV in the UK remained at low levels after the creation of devolved administrations in the late 1990s.
The trends are even more interesting when looking at RSV (Figure 3).Here, Italy clearly emerges as a country where this form of volatility has grown significantly, particularly after the 2008-2012 period.Spain recorded a slight increase during and immediately after the financial crisis and Great Recessionalbeit starting from higher RSV levels than Italyand then returned to greater stability.In Germany, region-specific volatility is effectively nonexistent throughout the three decades, while in the UK we see the opposite trend to Italy, with a clear stabilization of electoral dynamics.
It is already evident that all these complexities and nuances would not have been detected without the new volatility measures presented here.But we can go even deeper into the analysis of territorial variations by looking at what happens within each country, where we can assess the combinations of RTV and RSV and the existence of differences and similarities between individual regions.

Within-country variations
Italy: the triumph of multi-level instability.We have seen that, since the early 1990s, Italian regions have displayed the highest levels of electoral volatility.However, it is still unclear how RTV and RSV have combined in individual regions, i.e. what types of instability they have given rise to.Table 3 presents the four scenarios already considered in Tables 1 and 2 applied to 130 Italian regional elections.Here too there is a comparison between two periods of equal length: 1993-2007 (pre-crisis) and 2008-2022 (post-crisis).The percentages refer to the distribution of the four types of resulting stability/instability in the two periods.Already in the period 1993-2007, only 10.9% of regional elections were classified as stable (A).In the period 2008-2022, we do not find a single case in this category.Electoral instability thus continued and increased in the period 2008-2022, but with significantly different characteristics.In 1993-2007, the vast majority of regional elections were in quadrant B, indicating the existence of instability caused by electoral shifts between national (or macro-regional) parties.Therefore, the political earthquake that hit national politics also sent shockwaves through the lower levels of government.In the period 2008-2022, RTV levels remained high, but a further change occurred with the sharp increase in RSV.As a result, most elections moved into the D quadrant, indicating the emergence of multi-level instability.The high volatility of electoral support for national parties is now accompanied by increasing electoral fluidity in the regional sphere.This marks an important evolution in Italian regional politics, where change is no longer monopolized by state-wide parties, but also stems from sub-national developments.The multiplication of lists linked to regional presidential candidates and the emergence of various types of civic lists focusing on a single region have occupied an increasingly important place in Italian regional politics (Vampa, 2021a(Vampa, , 2021b)).This has taken place against a background of further destructuring of national politics.Figure 4 shows the electoral stability/instability of Italian regions in the periods 1993-2007 and 2008-2022.The initials of the regions (see appendix for full names) are distributed according to the average values of RSV (horizontal axis) and RTV (vertical axis) in each period.The grey ones with the circular markings represent the period 1993-2007, while the black ones with the triangular markings represent the later period 2008-2022.Clearly, we see a general shift of regions to the right (indicated by the dashed arrow), with an overall increase in region-specific volatility that is often accompanied by sustained levels of regiontranscending volatility.
Spain: diverging regional paths in a territorially fractured country.After the years of transition to democracy, Spain experienced a period of political stability that lasted until the start of the new millennium (Rodríguez-Teruel et al., 2018).This stability also extended to the regional level.Table 4 shows that almost two-thirds of Spanish regional elections in the period 1993-2007 (65.2%) were placed in quadrant A, indicating stability.The remaining elections were characterized by above-average levels of RSV resulting in regionally-driven instability (quadrant C), particularly in contexts where important regionalist parties were active.The period 2008-2022 saw a dramatic shift.Just over a third of regional elections still qualify as stable in the face of a marked increase in nationally-driven instability (quadrant B).In contrast to Italy, there are few cases of multi-level instability (quadrant D), while a substantial share of observations are still characterized by regionally-driven instability (quadrant C)the latter has declined but remains substantial.
Figure 5 shows how the distribution of regions (called 'autonomous communities' in Spain) changed from 1993-2007 to 2008-2022.In the first period, all regions are arranged in the lower part of the graph, indicating low RTV levels.There are regions, such as the Basque Country, Navarre, Cantabria and Canary Islands, placed to the right due to their higher RSV.The black triangular marks indicate the emergence of further divergence between regions in 2008-2022.The Basque Autonomous Community, Navarre and Canary Islands maintained above-average levels of RSV Table 3. Stability/instability in Italian regional elections (comparing distributions in 1993-2007 and 2008-2022).and other regions such as Catalonia, Galicia, Balearic Islands and Asturias moved to the right.Catalonia, in particular, suffered several region-specific electoral earthquakes in the post-2007 period due to the proindependence turn of previously moderate autonomist parties (Gray, 2020: 41-69).Almost all other regions, on the other hand, maintained low levels of RSV but shifted upwards to the left (growing RTV), experiencing nationally-driven instability.Madrid, Aragon, Castile and Leon, Andalusia and Murcia, for example, are regions where political change has been driven by the rise of new national political forces such as Podemos, Ciudadanos and Vox, the latter two advocating greater institutional centralization (Gray, 2020).In short, as indicated by the dashed line, two groups of regions seem to have emerged in the post-crisis scenario: those clearly moving towards nationally-driven instability and those maintaining or tending towards regionally-driven instability, among which Catalonia stands out as an important extreme case.

RSV
In contrast to Italy, the upper right-hand part of the graph (defining an area of multi-level instability) remains essentially empty.
Germany: the nationalization of instability.Germany experienced substantial political changes in the 1990s with the reunification process (Saalfeld, 2002).Five new Länder joined the Federal Republic and national politics saw the emergence of a far left party, the Party of Democratic Socialism (PDS), later transformed into the Left (Die Linke).However, the political scenario did not undergo any major shake-up in the period 1993-2007.
Table 5 shows that the vast majority of regional elections in this period (81.1%) qualified as stable (quadrant A).As already shown in Figures 2 and 3, RTV, which is underpinned by vote shifts between national parties, Table 4. Stability/instability in Spanish regional elections (comparing distributions in 1993-2007 and 2008-2022).

RSV Low High
RTV increased over time in German regions, while RSV remained very low.This led to the almost doubling of regional elections characterised by nationally-driven instability in the post-2008 period (31.5% in quadrant B, Table 5).In contrast to the period 1993-2007, this phenomenon now seems to affect the western and eastern regions of the country equally.Consistent with Table 5, Figure 6 shows two prevalent movements among the German Länder (with a few exceptions): one upwards and the other to the left.Länder that were very stable electorally in the 1990s and early 2000s, such as Baden-Württemberg, Rhineland Palatinate and Hesse, are now significantly more unstable, and this instability is almost exclusively determined by volatility between national parties.Bavaria has also moved to the topleft side of the graph, despite the growth of a Free Citizens list (Freie Wähler) in the 2000s, which seemed to suggest the emergence of region-specific electoral change (Vampa and Scantamburlo, 2021).In fact, the creation of regional and autonomous Free Citizens lists in various Länder has soon evolved into a process of political nationalization with the creation of an integrated federal party (also called Freie Wähler) competing in general and European elections (De Petris and Poguntke, 2015;Welsh, 2012).Similar nationalizing tendencies are evident in Bremen and Hamburg, which in the past, thanks to their city-state status, saw the emergence of small civic movements (their grey round markers are on the right of the graph).These weak local tendencies seem to have been absorbed into a framework of national political competition, which, at the same time, has clearly become more unstable in most regions.
The UK: consolidating regional politics.The UK differs from the other three countries in that the processes of regionalization and devolution did not affect the entire national territory.In the late 1990s devolved administrations were created in Scotland, Wales and Northern Ireland along with the Greater London Authority (Bogdanor, 2001;Mitchell, 2009).In these four cases regional representative institutionsparliaments or assemblieswere created.
The number of elections examined in Table 6 is much lower than in the other three countries but still gives us an overview of the stability/instability in British regional elections.We can see that only two of the four scenarios have occurred in the UK.In 1993-2007, the majority of elections for the four regional assemblies/parliaments were characterized by regionally-driven instability (quadrant C, 62.5%), the rest by stability (quadrant A, 37.5%).This is not too surprising given that, apart from London, the devolved administrations feature strong regionalist or even proindependence partiesand, in the case of Northern Ireland, the party system is almost completely regionalized.However, in the subsequent period (2008-2022) a general stabilization process seems to have taken place, with a majority of regional elections being in quadrant A (61.5%).
Figure 7 shows a general movement of regions to the left, indicating an overall reduction in region-specific volatility.For London, already characterized by very low RSV levels, the change is minimal.However, for Northern Ireland and Scotland, the shift to the left is sharp.In the case of Northern Ireland, the years between the late 1990s and early 2000s were characterized by major region-specific electoral transformations (McEvoy, 2008).In that period, the previously regionally-dominant partiesthe Ulster Unionist Party (UUP) and the SDLPwere replaced by the Democratic Unionist Party (DUP) and Sinn Féin (SF), which consolidated their primacy in the subsequent period.
In the case of Scotland, the period from the creation of the devolved administration to 2007 saw the gradual rise of the Scottish National Party (SNP) to the detriment of Labour (Mitchell et al., 2012) but also of other region-specific actors Table 5. Stability/instability in German regional elections (comparing distributions in 1993-2007 and 2008-2022).

RSV Low High
(such as the Scottish Socialist Party).Again, the years from 2008 to 2022 see the consolidation of SNP dominance with only a slight increase in RTV caused by vote shifts between Labour, Tories and Lib Dems.In Wales, the decrease in RSV is due to the stabilization of electoral support for Plaid Cymru (PC) and Labour's continued dominance (Davies and Wincott, 2022).In general, contrary to what we observed in the other three countries, UK regional politics (at least in the four cases of the first devolution wave) seems to have gone through a process of consolidation and stabilization.Even in the case of Northern Ireland, the growing tensions between the unionist and nationalist/republican camps do not seem to stem from increasing electoral volatility but, on the contrary, from an electoral scenario that, over time, has become more and more firmly and sharply divided between the DUP and SF.

Comparative discussion
The general picture of instability described at the beginning of this study becomes more complex and fragmented when considering the territorial roots of volatility in four major Western European countries.Indeed, Italy, Spain, Germany and the United Kingdom represent different trajectories of territorial political change that, although already hinted at in the literature, are more easily identifiable and quantitatively measurable thanks to the distinction between regionspecific and region-transcending volatility.
For example, if we had only considered the total volatility of regional elections in Spain and Italy, we could have simply concluded that a process of convergence towards instability occurred in the period 2008-2022.Indeed, the political landscapes of almost all regions in both countries have been increasingly subject to electoral change.However, the type of change has not been the same.
In Italy, the pre-existing high levels of regiontranscending volatility have been accompanied by a general increase in region-specific volatility.The result has been a territorial stratification of political instability that signals a process of decoupling between regional and national politics.National partiesirreparably damaged by the shock of the 1990s and in a constant state of organizational decline and regroupinglost control of key constituencies.This has led to substantial shifts of votes not only to other national challengers but also to region-specific parties controlled by autonomous local actors, whose role has strengthened visà-vis weakening national leaders (Grimaldi and Vercesi, 2018).
In Spain, on the other hand, we observe a division between autonomous communities subject to increasing levels of nationally-driven electoral change and others dominated by region-specific shifts.This division reflects the significant territorial tensions that have characterized Spanish politics in the post-crisis periodand which are not found in Italy (Vampa and Gray, 2020).Regions profoundly influenced by the rise of new parties with a strong national profile (including Ciudadanos and Vox) have contrasted with others more exposed to radical transformations in the regionalist camp.
Regional instability in Germany is in turn different from that in Italy and Spain and seems to be shaped by the centripetal nature of territorial relations in German cooperative federalism (Börzel, 2002).While in Italy we observed a multi-level disintegration of the political landscape and in Spain a polarization between different types of regional change, in the German Länder we witness a 'nationalization' of regional instability.In the period 1993-2007, only a few Ländermainly those in the Eastwere characterized by abnormal levels of volatility.However, in the subsequent period a general shift towards greater electoral fluidity reflecting national trends has taken place.In fact, new political organizations, often more rooted in certain regional contexts (such as the PDS or some extreme right-wing parties in the East of the country), were gradually absorbed into the national political landscape, leading to a territorial spread of fragmentation and volatility.Even when new, potentially region-specific actors such as the Free Citizens burst onto the political scene in some Länder (in Bavaria, for example), they soon went on to create a broader organizational network at the federal level that serves as a de facto national party.Thus, the German case shows that instability can coexist with a territorially integrated political system.
Finally, the UK differs from the other three countries due to the consolidation of its regional political systems in the post-2007 period.In Scotland, Wales, Northern Ireland and Greater London, the electoral fluidity that followed the creation of regional representative institutions in the late 1990s gave way to more stable political systems.Electoral stability, however, does not necessarily translate into institutional stability as it has exacerbated rigidities and fractures in intergovernmental relations between devolved authorities and central government (Esler, 2021) and between political communities within the same region in the case of Northern Ireland.
Table 7 offers a summary of the comparative analysis.The four countries (instead of individual regional elections) are placed in the conceptual framework of territorial stability/instability.While post-crisis Italy is in the upper right quadrant, the UK devolved administrations have moved to the lower left.Spain is split between regionallydriven (lower right) and nationally-driven (upper left) instability.Finally, Germany, placed in the upper left quadrant, represents a case of increasing electoral instability that almost entirely reflects national political dynamics.

Conclusion
This article has shown that the instability experienced by the political systems of various countries in recent years is a territorially complex phenomenon.It is not accurate enough to say that national and regional politics have moved in the same direction, just as it is not sufficient to automatically link the political volatility of one territorial level to that of another.Until now, great attention has been paid to the similarities/discrepancies between national and regional politics and how these are due to the existence of territorial political cleavages (around which regionalist parties are mobilized) or specific constitutional arrangements.However, analyses aimed at explaining vertical political integration rarely go beyond the small variations that characterize short-term electoral cycles.They tell us very little about how political change manifests itself in the regional political arena even in the absence of significant alterations in the politicization of regional identities or the level of institutional decentralization.This is also due to the fact that, so far, no measures of political-electoral instability have been proposed that take into account the territorial dimension and consider regional elections as democratic events in which nationalizing and regionalizing electoral pressures may coexist.By presenting the concepts of region-specific and region-transcending volatility (RSV and RTV) and applying them for the first time to 58 regions of four Western European countries, it was possible to observe that significant differences emerged between regions, countries and periods, thus going beyond 'static' accounts or the simplifying narrative of a general trend towards instability.The differences identified in the analysis above may in turn help us to describe and assess the divergent evolution of territorial politics in multi-level systems.
The conceptual development and application of the two types of volatility opens the way for new avenues of theoretical reflection and formulation of new hypotheses (to be tested by future studies).First, once the different types of volatility have been recognized and measured, their regional and national causes should be identified more rigorously.For example, to what extent is RSV influenced by the performance of regional institutions, regional economic growth (relative to the national economy) or the legacy of regionalist parties?Do party strategies play a role?Is RTV instead more sensitive to national political shocks than RSV?Is there a relationship between the two types of volatility?The Italian case seems to indicate that a prolonged period of high RTV, associated with a de-institutionalization of the national political landscape, may eventually result in a concomitant increase in RSV.This article has focused on four key European countries.However, the framework can be applied to any multi-level system in Europe and beyond.Moreover, it is possible to adopt a longer-term historical perspective, going beyond the last 30 years and comparisons based on the financial crisis as a critical juncture.Indeed, starting the analysis from an earlier point might reveal even more dramatic changes in the different types of volatility and resulting instability scenarios.After all, the early 1990s, with the end of the Cold War and the acceleration of the European integration process, were already a period of political turbulence compared to the previous two decadesas, once again, the Italian case illustrates well.
Moreover, while this article suggests the existence of a link between combinations of RTV/RSV and the broader evolution of territorial political dynamics in four countries (Table 7), it is still unclear whether it is regional instabilityin its different formsthat causes a shift in relations between central and peripheral policy makers or vice versa.One could also hypothesize the existence of a circularity.
Finallyand partly related to the previous pointwe need to assess the impact of different forms of instability on regional representation and governance and, ultimately, on regional policies.The main question is whether electoral volatility translates into changes of similar magnitude in the characteristics of those who lead regional institutions and in the policies they propose and implement.In short, the results presented here offer only the first insights into a new research agenda aimed at studying multi-level democratic systems in an era of increased political turmoil.

Table 1 .
Stability/instability at the regional level: four scenarios.

Table 7 .
The territorial dynamics of instability in four European countries: a summary of post-crisis scenarios.