Measuring evolving regional autonomy demands and statutes: introducing the Sub-state Autonomy Scale (SAS)

ABSTRACT This article introduces the Sub-state Autonomy Scale (SAS), an indicator and database for autonomy demands and statutes at the sub-state level. The usefulness of the SAS comes with two analytical properties. First, it allows evaluating autonomy in kind (whether competences are administrative or legislative), in degree (how much each dimension is present) and by competences (as a function of the extent of comprised policy domains) at the same time. Second, it can be used for analysing both autonomy demands and statutes. Taken together, the two aim at a refined analysis of sub-state mobilization and restructuring as well as their interplay.


INTRODUCTION
In the last century, many formerly centralized states considerably restructured their territorial organization by transferring political competences from the state to sub-state entities (Keating, 2017). In some instances, this decentralization followed functional ambitions to provide public goods at the most efficient scale. In others, it was the fruit of political contestation between the state and sub-state actors seeking recognition and power for regional political communities . Also, while in some cases the sub-state restructuring was limited to administrative decentralization, in others it went as far as to confer full legislative powers. The different regional political spaces that resulted from this variety of trajectories and powers, and its consequences for sub-state identities, institutions, interests and policies, have attracted significant attention from both regionalism scholars who examine the phenomenon itself (Allen & Cochrane, 2007;Aray, 2019) and comparatists who want to control for it (Loughlin et al., 2013).
To enable a systematic study of these questions, numerous efforts have been made to develop indicators accounting for the depth of decentralization statutes (Harguindéguy et al., 2021). More recently and fewer in number, indicators assessing the extent of decentralization demands by political parties have also been created. Despite their increasing in quality, there are two analytical properties that existing indicators do not yet comprise. First, none allows one to assess decentralization in kind (i.e., distinguish between different types of autonomy), in degree (i.e., distinguish between different extents of autonomy) and by competences (i.e., for various policy fields) at the same time. Second, they concern either decentralization statutes or demands, but there is no indicator available for both.
When considering that differences in power and in political dynamics come not only with gradual but also with qualitative (administrative versus legislative) autonomy differences as well as with the depth of controlled policy fields; and when considering that the interplay between autonomy demands and statutes provides important insights for the evolution of sub-state restructuring (McLean, 2001), there is a clear need for an indicator to incorporate these properties when studying these issues. Based on this observation (coming with the needs of my own research), I created the Sub-state Autonomy Scale (SAS), a 0-1 ratio relying on three dimensions of autonomy (group rights, self-rule and shared rule) that allows for the evaluation of sub-state autonomy demands and statutes in kind, in degree and by competences at the same time. It is suitable for regression, set-theoretic and qualitative analyses. At this stage, it is available for sub-state entities in Belgium, Spain and the UK between 1707 and 2020 (starting dates vary across countries, see below). Updates and extensions are possible.
The remainder of the article is structured as follows. The next section is dedicated to the review of existing indicators of sub-state autonomy statutes and demands. In the third section, the conception of the SAS is introduced, together with its database. The fourth section illustrates how the SAS measures the evolution of autonomy in different cases compared to existing indicators and how it allows to directly compare autonomy demands and statutes in a unified framework. The article concludes with a reflection on the main strengths and limitations of the SAS, as well as on the perspectives it provides for the study of contemporary sub-state mobilization and restructuring.

MEASUREMENTS OF SUB-STATE AUTONOMY
The increasing interest in measuring sub-state autonomy statutes and demands comes with two developments. The last 50 years have witnessed an increasing demand and conferral of political autonomy below the state level, which was followed by an equally increasing scholarly interest in understanding the determinants and consequences of the visibly varying types and degrees of demanded and enacted political decentralization (Keating, 2017). At the same time, political science has become increasingly comparative and experienced a revived interest for (regional) institutions and party politics since the 2000s (Erk & Swenden, 2010;Hepburn, 2009).
Taken together, both developments led to the creation of about 25 decentralization indicators that inspired Harguindéguy et al. (2021) to provide a comprehensive review. More recently and fewer in number, scholars interested in party politics started to create indicators that assess the type and extent of autonomy demands by national or regional(ist) parties. This section reviews those that measure the type or degree of regional (not local) autonomy with a generalist approach (i.e., accounting for more than one policy field). Doing so, it explains why a new indicatormeasuring autonomy not only in kind, in degree and by competences at the same time but also allowing to directly compare autonomy demands and statutes in a unified frameworkis needed.

Existing indicators measuring autonomy statutes and demands
For autonomy statutes, 10 indicators measuring the type or degree of regional autonomy with a generalist approach were identified, as listed in Table 1. When comparing them, four main lessons can be drawn. First, excepting the Regional Authority Index (RAI), which analyses regional tiers of government, all indicators are measured at the country level (with the Autonomy Index (AI) accounting for regional asymmetry). Second, while most measure autonomy foremost in degree (those by Lijphart, 1999, and Treisman, 2002, could also be used as categorical in kind measurements), the Unitarism-Federalism Scale and AI are the only ones to assess autonomy in both degree and kind at the same time. Third, the RAI and AI are the only indicators that try to account for varying degrees of competences. Fourth, the RAI stands out in both cross-sectional (728 regions/tiers) and longitudinal comparison (68 annual measurements).
For autonomy demands, three indicators assessing the type or degree of autonomy demanded by national or regional political parties were identified, also listed in Table 1: the Chapel Hill Expert Survey, the Ethnopolitics and Party Competition dataset, and the Regionalist Party Ideology Scale. When comparing these, three more points stand out. First, all indicators are party-level studies, which is probably the most suitable approach given the heterogenous agency of sub-state entities and valid constructivist warnings against essentialization (Paasi & Metzger, 2017). Second, while the Chapel Hill Expert Survey and Ethnopolitics and Party Competition dataset scales measure demands in degree, the Regionalist Party Ideology Scale relies on a typology with in-kind distinctions. Third, while the former two stand out in cross-sectional comparison (22/31 countries), the latter does so in longitudinal (56 years).
With the help of these measurements, scholars were able to identify the overall trajectories, logics and structural reasons for de-and recentralization in many regions of the world from the middle of the last century (Arzaghi & Henderson, 2005;Dardanelli et al., 2018;Lane & Ersson, 1987/1999Lijphart, 1999Lijphart, /2012Shair-Rosenfield et al., 2021;Treisman, 2002;Woldendorp et al., 2000). The measurements also made it possible to look into the consequences of decentralization, for example, on policy performance (Keman, 2000), economic growth (AER & BAK, 2009;Filippetti & Sacchi, 2016) or corruption (Neudorfer & Neudorfer, 2015). The measurement of autonomy demands, in turn, was associated with that of autonomy statutes when explaining how decentralization impacts regionalist party strength (Brancati, 2006;Massetti & Schakel, 2017) and ideological radicalization (Massetti & Schakel, 2016). It was studied for itself when showing how regionalist party positions vary based on ethnic voting, ethnic party competition and the relative importance of different policy issues .
Why another indicator for measuring autonomy statutes and demands? When looking at the decentralization processes that many formerly centralized states have undergone in the last 50 years, changes in kind (i.e., distinguishing between different types of autonomy), in degree (i.e., distinguishing between different extents of autonomy) and by competences (i.e., for various policy fields) appear of considerable political importance. In Belgium, for example, federalism and the legislative competences that came with it have been the 'f-word' of politics from their first mention in the 1900s until after the first state reforms in the 1970s and 1980s. Since then, the country has seen five more heavily debated state reforms with incremental competence enlargements in fewer than 45 years (Deschouwer, 2012). 'Devolution' of legislative powers in the UK was Note: For an approach measuring 'territorial political capacity' rather than autonomy or authority, see Cole et al. (2018). For data on federal entities' policy autonomy in various fields without operationalization as indicator, see Dardanelli et al. (2018). For a dataset with regionalist parties' position on various types of territorial demands and their framing without operationalization as indicator, see Elias et al. (2021). Sources: Schakel (2008), Dardanelli (2019), Harguindéguy et al. (2021) and author's own research for autonomy statutes; author's own research for autonomy demands.
equally contested from the 1920s to the 1990seven within Scotland and Wales. After their statutes had been in place, they were extended twice for both of them following regionalist pressures (Deacon, 2012). In Spain, it was only after important regional mobilization in the 1970s and 1980s that autonomy statutes with legislative powers were envisioned for entities other than the so-called 'historic' communities of Catalonia, the Basque Country or Galicia. Once again, two competence extensions have fuelled political discussions (Moreno, 2002). In centralized cases such as metropolitan France, in turn, demands for regional legislative autonomy have been repeatedly rejected until todayeven when faced with violent upheavals such as in Corsica, whose assembly's acts still need national parliamentary assent to acquire legal force (Kernalegenn & Pasquier, 2018). One could go on with this illustration, but two points should become clear. First, decentralizing administrative competences and decentralizing legislative competences are very different things. With legislative institutions come sub-state policies that can shape different sub-state interests and identities. Second, once legislative institutions are in place, the distribution of competences between the state and sub-state level is of critical importance and can lead to considerable tensions between both levels (Toubeau & Wagner, 2015).
The two existing indicators of sub-state autonomy statutes that stood out in the preceding review are the RAI and AI. Yet even these most advanced indicators do not, despite their merits pointed out below, fully allow the capture of the political dynamics described above.
While the RAI provides an impressively detailed additive account of a region's degree of decentralization on 30 points, it does not specify their relative position vis-à-vis a qualitative anchor that would allow to distinguish administrative from legislative powers. All self-rule components (institutional depth, policy scope, fiscal autonomy, borrowing autonomy and representation) are indeed added up irrespective of regions' legislative capacity (Hooghe et al., 2016, p. 106). This leads, for example, to entities with administrative competences such as the French regions  to score 10 on self-rule, while entities with legislative competences such as the Flemish or Walloon regions  in Belgium score 7 and 6, respectively. The French score is higher because of the fiscal and borrowing autonomy of the regions, which were at that time absent in Belgium. Yet, the Belgian regions enjoyed legislative competences in different policy fields. Now, one may just want to operationalize things differently based on the RAI's policy scope dimension that distinguishes between 'very week authoritative competences' (0) and 'authoritative competences' in several extending policy sectors (1-3) (p. 28). By 'authoritative', the authors mean 'having the capacity to develop binding rules through legislation or executive orders' (p. 67). However, this does not capture the difference between administrative and legislative competences because the French regions (1986-2010) obtain the same score (2) as Flanders (1980-88), while Wallonia (1980-88) even gets a smaller one (1).
When we look at the way the RAI captures competence enlargements, we see that the policy scope dimension distinguishes between economic policy, cultural-educational policy, welfare policy and special competences such as residual powers, police, own institutional set-up and local government. Score 1 is given for 'authoritative' competences in one of the above. Score 2 is given for two of the above. Score 3 is given for special competences and two others of the above. Score 4 is given when, furthermore, there are competences over immigration, citizenship and right of domicile. While this covers an impressive variety of competences, one should note that having some competences in one domain qualifies as being competent in it. In other words, there is no account for how much decentralization in each domain occurred. This leads entities such as the Spanish autonomous communities  to score 3 for their policy scope, although most of their competences are only partial. Furthermore, the 1992 extension remains unnoticed because there were pre-existing competences in the extended fields (education and health care, see Machado, 1992). An entity such as Scotland (1999-2010) equally scores 3 on the policy scope dimension, although its devolved policy areas were much more comprehensive than the Spanish ones (Cairney, 2011).
The AI, in turn, varies a little less (14 points, with 7 additional half-points) but is conceived in such a way that the point-8 anchor distinguishes between administrative and legislative autonomy. In addition, when assessing policy autonomy (0-7 points), it accounts for whether substate entities have concurrent or exclusive competences in one or at least two 'key areas of government' (economic policy, cultural-educational policy, welfare state policy) (Dardanelli, 2019, pp. 280, 285). Concurrent primary law-making powers are worth 3.5 points when extending to one area. They are worth 4 when extending to two. Exclusive primary law-making powers are worth 5 points when extending to one area. They are worth 6 when extending to two. The score is further increased by half a point when sub-state entities have control over either civil and criminal law, police, the judiciary or local government, or when they have a veto on central government decisions affecting the entity. It is decreased by half a point when the central government can veto sub-state decisions. While this constitutes a considerable advancement towards a more accurate assessment of the extent of competences held by sub-state entities, extensions are not captured if a sub-state entity already possesses competences in two of the three areas. For example, the Spanish autonomous communities score 4.5 on the policy autonomy dimension from 1985 to 2015without an increase between 1990 and 1995 for the 1992 reform mentioned above. In Belgium, the federal entities score 6.5 on the policy autonomy dimension from 2010 to 2015without an increase for the sixth state reform in 2014 which conferred further competences in the social security sector. Two more caveats deserve mentioning. On the one hand, the coding is expressed in quinquennial intervals, which makes it more difficult to account for their precise historical occurrence. In addition, given that areas are frequently subject to shared competences, the comparison of state and sub-state competences in a particular policy field could be further refined.
For indicators of sub-state autonomy demands, although significant progress has been made in recent years, existing measurements are less sophisticated than for autonomy statutes. While the Regionalist Party Ideology Scale comprises four in-kind differentiations, the Chapel Hill Expert Survey question and the Ethnopolitics and Party Competition scales provide relative in degree differentiations, but none combines both in kind and in degree differentiations or explicitly looks for particular policy fields.
Finally, one should note that there is hitherto no decentralization indicator with data available for both autonomy demands and statutes. Yet, it has been shown that decentralization reforms have often been enacted following political pressures by regionalist parties (Hepburn, 2009;Toubeau, 2018), while decentralization has, at the same time, the potential to create new demands (Erk & Anderson, 2009). To disentangle these dynamics, research on sub-state mobilization and restructuring would clearly benefit from an indicator available for both autonomy demands and statutes. Now, to avoid any misunderstanding, the argument is not that existing indicators are flawed or henceforth uselesson the contrary. That the RAI and AI do not account for all competence extensions comes with the advantages of parsimony if their degree of precision is sufficient for specific research (especially when needing a broad cross-sectional coverage such as that of the RAI). Furthermore, that the RAI and the AI do not account for autonomy demands, and that the three demand indicators do not account for autonomy statutes, comes with the fact that they were not designed to do so. The argument is that research problems that look into political dynamics that are influenced (1) by the qualitative distinction between administrative and legislative autonomy and by competence distributions and extensions; or (2) by the interplay of autonomy demands and statutes, can be resolved more meaningfully with an indicator such as the SAS.

Conception
The conception of the SAS required three conceptual and methodological steps, presented hereunder: (1) defining the concept of sub-state autonomy; (2) operationalizing its dimensions; and (3) translating the operationalized dimensions into a scale measuring the defined concept.

Concept
The SAS aims to capture what prerogatives sub-state entities demand and obtain to become more 'autonomous'. Etymologically, 'autonomy' means living according to one's own rules (Greek: 'auto' meaning 'self; 'nomos' meaning 'rule'/'law'). Therefore, sub-state autonomy is defined here as the 'legal prerogatives allowing a political entity below the state level to live according to its own rules'.
Three types of prerogatives allow a political entity to become more autonomous: group rights, shared rule and self-rule (Barter, 2018). Group rights allow a political community to preserve its culture and use its language in the main state institutions (e.g., administration, education, judiciary)provided that these are different from the rest of society. Shared rule allows an entity to co-determine decisions taken for the state as a whole. Self-rule prerogatives, finally, allow an entity to take decisions for itself alone.
Depending on how broadly autonomy is conceived, the prerogatives that are associated with it differ. A narrow approach, such as the one Dardanelli (2019) uses for his AI, only includes self-rule because it is the only prerogative allowing a sub-state entity to take decisions on their own. A broader approach recognizes that shared rule allows (co-)determination of the rules according to which one lives and should be added to self-rule. This is what  do with their RAI. The present approach goes even further by considering that decisions taken outside the sub-state entity can also render it more autonomous when allowing a political community to integrate society while preserving its own language and culture. For the SAS, autonomy is hence conceived as encompassing self-rule, shared rule and group rights. Despite this broader approach, it needs to be taken into consideration that these dimensions weight differently in rendering an entity autonomous: with self-rule prevailing over shared rule, which prevails over group rights. This reflects in the scaling below.

Operationalization
Group rights can be operationalized based on different prerogatives with gradually increasing importance. Being recognized by the state (in law or practice) as a relevant political group (not only as an administrative division) is one thing. Being allowed to use an own different language in primary education, secondary education, administrations, the judiciary and supporting cultural distinctiveness is another. For the SAS, when one to three of the former prerogatives are involved, cultural and linguistic facilities are considered to be partial. When four to five are involved, cultural and linguistic facilities are considered to be advanced. When sub-state entities do not have such specificities, this dimension is only of limited relevance and not taken into consideration.
Shared rule, in turn, can be operationalized based on co-decision power in different institutions. The SAS accounts for the two most important ones, namely the state parliament and the executive. Since degrees of influence within these institutions can vary, whether sub-state entities only have a guaranteed representation in them or if their support is also required to take decisions, that is, if they have veto powers, is accounted for.
Self-rule, finally, involves a variety of operationalizable sub-dimensions. Its depth is determined by the capacities of institutions and their policy scope. For institutional capacities, one can rely on the distinction between 'administrative' and 'legal' self-rule.
Administrative self-rule allows sub-state entities to implement policies whose content has been determined by state institutions. Three types with increasing importance exist. (1) Executive administration allows the implementation of policies through a sub-state ministry, but as determined by state institutions. (2) Advisory administration, in addition, allows for the voicing of sub-state concerns through an own representative body.
(3) Discretionary administration allows sub-state entities to decide themselves how they implement predetermined policies. Furthermore, administrative autonomy increases in function with the controlled policy areas, with fiscal and borrowing autonomy and with institutions' legitimacy. While executive administration relies only on an own ministry, advisory and discretionary self-rule can come with an elected assembly that deepens the popular ownership. Discretionary administration can even involve an own executive.
Legislative self-rule, in turn, allows to determine directly the content of specific policies. Its scope increases with four features. (1) Institutional legitimacy (own assembly and executive) deepens popular ownership. (2) Controlled policy fields determine institutions' room for manoeuvre. The SAS accounts for: culture; education; social and sanitary affairs; territorial affairs; economy; social security; citizenship and immigration; and five special areas of government (local government, police, defence, penal and civil law, judicial system). Table A1 in the supplemental data online comprises their detailed content. (3) Fiscal and borrowing autonomy determine how independent from state resources a sub-state entity is. (4) Statutory autonomy determines how autonomous sub-state institutions themselves are: whether their functioning can be regulated, whether sub-state laws are subject to state veto, whether the statute is constitutionally guaranteed and whether international relations in own competence areas can be conducted.

Scaling
Based on this conceptualization, the SAS was conceived as a 0-1 range following three principles (Tables 2 and 3). First, based on the hierarchy elaborated above, 0-9% of the score was attributed to group rights, 0-16% was attributed to shared rule and 0-84% was attributed to self-rule. The sum does not equal 100 because self-rule and shared rule can be assessed independently of group rights for entities without cultural or linguistic distinctiveness. This means that group rights provide a higher SAS score in the absence of self-rule prerogatives. Once disposing of self-rule prerogatives, the SAS score follows fixed anchors on the self-rule dimension (see below). Like this, group rights prerogatives, which are often conferred prior to self-rule, are captured, but two sub-state entities with identical self-rule prerogatives will obtain the same score regardless of their group rights. Both sub-state entities with and without cultural or linguistic specificities can thus be directly compared. The shared rule score is always added up, either to the group rights score if the sub-state entity does not have self-rule prerogatives, or to the selfrule prerogatives if present (Table 3).
Within legislative self-rule, the largest score share is attributed to controlled policy fields and special government areas (59%) because they are most significant in determining its depth. Institutional legitimacy, fiscal and borrowing autonomy and statutory autonomy count each for 12-18% of the share. While these differences aim at respecting the conceptual hierarchy elaborated above and the relative importance between groups of subdimensions, the precise shares were derived after giving equal weight to all components within a sub-dimension. Among policy fields and special government areas, no weight differences were introduced to remain value-free. Since the data are coded separately for each, weights can be given in specific research if deemed appropriate.
Second, the scale relies on additive in-degree measurements within fixed scores defined by qualitative anchors on the self-rule dimension. The main anchor distinguishing between administrative and legislative self-rule is set at 0.50 to have a central anchor point. The anchors for the sub-dimensions of administrative self-rule are set at 0.12, 0.16 and 0.22 based on equal weighting of their components. To respect the 0.50 anchor, shared rule is added either to administrative or to legislative self-rule so that the total score never lies above 0.50 with administrative selfrule and never lies below 0.50 with legislative self-rule.
Third, in-degree scores are assigned based on the simple presence or absence of the listed prerogatives, with two exceptions. (1) For policy fields, special government areas and statutory autonomy, sub-state entities' competences are assessed relative to state competences. If an entity has exclusive competences, it receives the full score for a domain; if an entity has more competences than the state, it receives 75% of the score; if the entity and state competences are about equal, 50% of the score is given; if an entity has fewer competences than the state, it receives 25% of the score; if an entity has no competence, it receives 0% of the score. (2) For fiscal and borrowing autonomy, the SAS follows the RAI conceptualization (while subjecting it to the aforementioned anchors). Concerning fiscal autonomy, 25% of the score is attributed if the rate of minor taxes can be set; 50% is attributed if the base and rate of minor taxes can be set; 75% is attributed if the rate of one major tax (personal income, corporate, value added or sales) can be set; the full score is attributed if the rate and base of one major tax can be set. Concerning borrowing autonomy, 33% of the score is attributed if borrowing requires authorization by the state; 67% is attributed if only certain state rules apply; the full score is attributed if there are no restrictions. Notice that fiscal and borrowing autonomy are taken into consideration for both administrative and legislative self-rule because both can entail them. Fiscal and borrowing powers are considered as administrative when they consist in administrative acts in application of state laws. They are considered as legislative when they are exercised through own sub-state laws. Finally, for independence demands (not sub-state but state autonomy), the score 'Ind.' was assigned. The conception, operationalization and scaling of the SAS is summarized in Tables 2 and 3. While Table 2 lists all sub-dimensions and has the score cumulatively following the scale, Table 3 shows how real case calculations are made.

Database
After conceiving the scale, I created a database with SAS scores for the autonomy statutes and demands of sub-state entities in Belgium, Spain and the UK (coming again with the needs of my own research). This involved three more conceptual and methodological choices, clarified hereunder: (1) defining and operationalizing autonomy statutes and demands; (2) defining and selecting the sub-state entities under study; and (3) collecting and verifying the data for them.

Operationalizing autonomy statutes and demands
To assess the 'autonomy statutes' of sub-state entities, the legal prerogatives obtained by them were taken into consideration. Assessing their 'autonomy demands', in turn, is more complicated because various actors (e.g., political parties, social movements, individuals) can issue them. Since political parties campaign with autonomy demands (or against them) in elections, can hence make certain claims of representativeness and are ultimately those who negotiate autonomy statutes with state elites, the present approach focusses on the autonomy demands of political parties within sub-state entities. Now, such autonomy demands can also be driven by different parties. With a high number, however, it is not only time consuming but also conceptually difficult to assess all parties' positions precisely and coherently. To achieve such precision, and since there are two types of parties that can be considered particularly relevant for the autonomy dynamics in sub-state entities, the present approach only focuses on two following types of parties in each studied sub-state entity.
Regionalist parties, that is, parties whose foundational issue is the defence of an entity's political interest and/or autonomy, are usually pictured by the literature as reference for the autonomy demands in sub-state entities (Hepburn, 2009). Since there can be several regionalist parties within a sub-state entity, the one with the highest electoral support is retained here because it can be deemed most representative for an entity's regionalist voters and most relevant for the perception of its regionalist demands by external actors. Before being considered relevant, a regionalist party needs to obtain at least 5% in regional or national elections.
Traditional parties, that is, parties that compete on the original political cleavages and are usually those detaining power in government, can also be decisive drivers of the autonomy question (the term 'traditional' is not related to their age but to them being in power and following one of the main traditional liberal, socialist or centrist positions). Although they are usually seen as the competitors and as threatened by regionalist parties, some traditional parties take up the autonomist question (because of this threat and/or as a political strategy) and become defenders of moderate forms of sub-state autonomy (Toubeau, 2018). While several traditional parties can have a positive attitude towards the autonomy question, there is often one pushing the most and strategically portraying itself as the main non-regionalist defender of autonomy. Examples today would be the Christian-Democrat party in Flanders, the Socialist Party in Catalonia or the Labour Party in Scotland. For the traditional party demand within a sub-state entity, it is therefore the official position of the traditional party with the highest autonomy demand that is taken into consideration. 1 If two (or more) traditional parties have an identical position, this common position is taken into consideration. If the autonomy demands of one traditional party overtake another, the latter is substituted with the former. For example, in Scotland the Labour Party has been the most visible traditional party pushing for devolution since 1974. However, its position was not explicit on the matter between 1934 and 1973, unlike that of the Liberal Party, which I take as reference for this period. The objective of this approach is to continuously capture the highest degree of autonomy demanded by one or several traditional parties within a sub-state entity at a given point in time. This comes at the cost, however, of not always referring to the same party behind the demands, and of doing so regardless of traditional parties' vote share. Should the latter be important for particular research, then the demanded degree of autonomy could be weighted or put into relation to the parties' vote share.

Case definition and selection
By transposing Weber's (1919Weber's ( /2004 conception of the state to the sub-state level, sub-state entities can be conceived in three different ways: as an institution applicable to the inhabitants of a certain territory, as a territory inhabited by a human community with or without special institutions, or as a human community inhabiting a certain territory with or without special institutions. The SAS follows the latter approach by calling these entities 'sub-state communities' and by defining them as 'social groups for which autonomous institutions have been demanded or to which autonomous institutions have been conferred on a specific territory'. The objective of this conception is to do justice to the sociological dynamics of sub-state mobilization and restructuring, while at the same time avoiding essentializing sub-state communities as inherent and uniform social actors (Paasi & Metzger, 2017). By acknowledging the constructed nature of social groups and identifying them either based on an institutional reality or a significant political mobilization, the definition follows what Zuber (2013) developed as a 'naturalized constructivist approach'. It is constructivist because it conceives social groups as formed by social practice and cognition rather than as intrinsic and essentialized categories. It is naturalized because it considers social groups as meaningful categories of analysis since they appear to be an important frame of reference for the agents of sub-state mobilization and restructuring themselves. As was shown in the operationalization of autonomy demands above, it is not assumed that all the inhabitants of the territory for which the autonomy is claimed or to which it is conferred subscribe to the asserted political identity or autonomy claim.
Note that this conception is anachronistic insofar as institutional or political communities today did not always exist in the same form in the past. The objective is to study institutional and political change longitudinally by accounting for the before and after of this change. The historical significance of the data needs to be interpreted accordingly. Table 4 lists all sub-state communities that were retained in the three countries. Table 3. Real case calculations for the Sub-state Autonomy Scale (SAS). If sub-state autonomy demands or statutes include … … they score legislative self-rule 0.52 + its scores on this sub-dimension + its scores on the shared rule dimension ¼ 0.52 (minimum) -1.00 (maximum) discretionary administrative self-rule 0.22 + its scores on this sub-dimension + its scores on the shared rule dimension ¼ 0.22 (minimum) -0.48 (maximum) advisory administrative self-rule 0.16 + its scores on this sub-dimension + its scores on the shared rule dimension ¼ 0.16 (minimum) -0.36 (maximum) executive administrative self-rule 0.12 + its scores on this sub-dimension + its scores on the shared rule dimension ¼ 0.12 (minimum) -0.30 (maximum) no self-rule 0.00 + its scores on this sub-dimension + its scores on the shared rule dimension ¼ 0.00 (minimum) -0.25 (maximum)

Data collection and verification
The starting point for the data collection is the modern period of states' history, provided that communities already belonged to it (otherwise it starts with this belonging). For Belgium, the observation starts with the creation of the state in 1830. For the UK, it starts with the Union of England and Scotland in 1707. For Spain, it starts with the first transition to democracy after the 1868 Glorious Revolution. Observations end in 2020. Communities within contested territorial spaces such as Brussels or Northern Ireland were analysed twiceonce for each political community in the territory. Further country studies and updates are possible.
To collect the necessary information for attributing SAS scores, two types of sources were used. First, I read political, historical, legal and sociological secondary literature and assigned scores provisionally on this basis. Cases were disaggregated temporarily for each change in either traditional party demands, regionalist party demands or the autonomy statute of a community. Ranges were expressed annually. Second, I conducted expert interviews to discuss and verify all scores. 2 For each country, I started interviewing legal scholars to check the coding for all communities' autonomy statutes. I continued by interviewing political scientists and historians on regionalist and traditional party demands to verify my coding for themensuring as main coder both coherence within and across cases. In case of doubts, I consulted additional experts until I had sufficient information to assign the scores with confidence. In total, I conducted 44 interviews. Eventually, I compiled a database with scores for all scale components so that aggregated values can be verified and adapted for specific research purposes. The database is available and will be maintained in case of updates or extensions on the following dataverse: https://doi.org/ 10.34934/DVN/LSXXZV.
To ensure scientific quality, the process was guided by four criteria, as set out by Harguindéguy et al. (2021). The scale conception and data collection were (1) presented transparently in this article; and (2) made available publicly. (3) Two validity checks were conducted: conceptual validity through critical exchange with other scholars, and content validity through systematic verification of the coding by case experts. Convergent validity with existing measurements, although checked and confirmed in Figures A1 and A2 in the supplemental data online, was not retained as a validity criterion because, despite significant intercorrelation, substantial conceptual and empirical differences between the measurements exist. They are illustrated in the next section. Finally, (4) reliability was sought through 'snowball checks' with unconfirmed information being (re)submitted to experts until confirmation could be given. The present approach follows Hooghe et al. (2016, pp. 30-32) in not conducting intercoder reliability tests because for complex measurements such as the RAI or SAS, dialogue with experts is essential and can be considered much more reliable than a non-dialogued expert survey.

ILLUSTRATION AND COMPARISON WITH EXISTING INDICATORS
In this last section, the usefulness of the SAS will be illustrated vis-à-vis the two main properties that justified its creation: (1) the need for an indicator that assesses autonomy in kind, in degree and by competences; and (2) the need for a unified framework when studying autonomy demands and statutes. Put differently, the objective is to show that the scale lives up to the raised expectations and, at the same time, to provide some case illustrations. Assessing autonomy in kind, in degree and by competences To demonstrate the capacity of the SAS to measure autonomy in kind, in degree and by competences at the same time, and that this actually yields different results than existing indicators, the illustration starts with a discussion of the Flemish case by comparing the entity's scores on the SAS, RAI and AI. The starting point is 1950, that is, when the first AI data are available. The important evolutions that one would want to see captured are the extension of cultural and linguistic rights to an own executive administration in 1964, the conferral of legislative institutions with a first set of competences through the first state reform in 1970, and the subsequent competence extensions through the second (1980), third (1989), fourth (1993), fifth (2001) and sixth (2014) state reforms. As shown in Table 5, all these changes are captured by the SAS that accounts for the shift from administrative to legislative autonomy in 1971 with the score passing from below to above 0.50, as well as for the extensions of both administrative and legislative autonomy with extending scores. The RAI (self-rule dimension) and AI, in turn, account for some of these changes but not for all. The 1990 competence extension remains unnoticed by the AI because there had been some decentralization already in the extended fields (educational and social and sanitary matters). The 2014 competence extension remains unnoticed by both the AI and RAI because Flanders was already competent in more than two important areas of government they account for. While the shift from administrative to legislative autonomy is the starting point for the RAI measurement and captured by the AI with a score increase, both fail to distinguish it from entities with administrative powers. For the RAI where administrative entities can score higher (see discussion above), this is due to the absence of a fixed anchor distinguishing administrative from legislative powers. For the AI, this is probably due to a coding mistake because it comprises such a fixed anchor (8) and it works in other cases (see below). As for the linguistic partition of the national ministries for culture and education in 1964 that conferred de facto an executive administrative autonomy to the two main linguistic communities, it also remains unnoticed by the AI.
Second, the illustration is pursued with the case of Scotland that has, similarly to the Flemish case, obtained legislative autonomy after a long period of political mobilization but with far fewer incremental competence increases. Instead, it received a higher degree of autonomy at the beginning, with fewer and less far-reaching subsequent extensions than Flanders. The autonomy changes one would like to see captured are the shift from the administrative Scottish Office to the legislative Scottish Parliament in 1998, as well as the competence extensions of the 2012 and 2016 Scotland Acts.
As shown in Table 6, the SAS is again able to capture all these changes, both with regard to the qualitative turnover point discriminating administrative from legislative autonomy and with regard to its subsequent Table 5. Regional Authority Index (RAI) (self-rule), Autonomy Index (AI) and Sub-state Autonomy Scale (SAS) scores for Flanders. extensions. The RAI and AI, in turn, also register a considerable increase in 1998, which this time correctly crosses the legislative anchor for the AI. They fail, however, to account for the following competence extensions (keeping in mind that the AI observations stop in 2015). This is once again due to the fact that important political competences had been conferred in more than two policy fields previously. The problem of not registering any competence extension is more pronounced than in Flanders, though, because in Scotland higher levels of sub-state autonomy had been immediately conferred in 1998. Besides correctly capturing these evolutions, the SAS also provides coherent absolute scores with the Flemish statute exceeding the Scottish. This comes mainly with its larger economic competences, borrowing autonomy, international relations capacities and some slightly exceeding areas (culture, education, local government), while the Scottish statue only exceeds the Flemish in a few areas (police, judicial system, penal/civil law, constitutional autonomy) (see detailed scores in the database).
A unified framework for studying autonomy demands and statutes To illustrate the ability of the SAS to provide a unified framework for studying the evolution of autonomy demands and statutes, another example is used, that of Catalonia. Since a detailed discussion is required, the illustration is limited to one case. In the absence of an existing indicator available for both autonomy demands and statutes, only the SAS is taken into consideration.
In Catalonia, whose SAS scores are summarized in Table 7, the first demands for sub-state autonomy came up in 1871 with the Partido Republicano Democrático Federal, a wing of the traditional republican party, in favour of a federalized Spain. It lost significance during the arranged alternation in power (Turnismo) established by the Conservative and Liberal parties between 1876 and 1906, but came up again in 1907 when a massive vote for other parties started to end the Turnismo in Catalonia. The regionalist Lliga Regionalista was founded with an even more far-reaching federalist demand. In 1913, some symbolic recognition of the Catalan demands was granted This short historical revisiting of the autonomy dynamics in Catalonia since the late nineteenth century, reflected in the SAS scores in Table 7, illustrates how traditional party demands, regionalist party demands and autonomy statutes evolved in this sub-state community. While unfortunately not all sub-dimensions and fields behind the scores can be discussed (for this, see the database), the illustration gives a sense of how the SAS allows to trace these evolutions, to directly compare autonomy demands and statutes on a single indicator and, by doing so, provides a unified framework for the analysis of autonomy demands and statutes.

CONCLUSIONS
In recent decades, numerous states around the world have entrusted sub-state entities with increasing political autonomy. In many of them, this process has been pushed forward through the interplay of sub-state actors demanding and state actors conferring autonomy (Keating, 2017). The determinants and consequences of these phenomena incited a revived interest in regional institutions and party politics (Erk & Swenden, 2010;Hepburn, 2009) which, together with a comparatist turn in political science, led to the development of numerous indicators trying to measure both the extent of autonomy demands and statutes.
When looking at the political processes behind evolving autonomy demands and statutes, differences in autonomy types (administrative versus legislative), degrees (different prerogatives) and competences (policy fields) have proven of significant political salience. To account for these dynamics and issues, an indicator needs to (1) measure sub-state autonomy in kind, in degree and by competences at the same time; and (2) be available for both autonomy statutes and demands. In the absence of an existing indicator capable of fulfilling them, these two properties motivated and guided the creation of the SAS, introduced in this article.
Notwithstanding these merits, some limitations need to be signalled. First, the detailed nature of the SAS comes at the cost of a complex and time-consuming data collection process. Consequently, despite further temporal updates and country extensions being possible, its cross-sectional coverage is lower than that of the most relevant existing indicators. Second, by giving priority to a precise account of legislative self-rule whose evolution is crucial for the aforementioned dynamics, the SAS is less detailed on the administrative self-rule dimension than the AI, and less detailed on the shared rule dimension than the RAI. Third, while the relative weight of the main sub-dimensions followed conceptual considerations of their relative importance, the weight of elements within these subdimensions were kept constant. Since the importance of elements such as policy competences might vary across context and time, this islike in every scalesubject to some arbitrariness. Fourth, assessing political parties' autonomy demands comes with the difficulty of determining the positions of heterogenous and numerous actors. Systematic expert checks were conducted to determine parties' positions as precisely as possible. Their large number, however, made it possible to consider only the demands of the two most relevant autonomy actors in the party spectrumthe regionalist party with the highest electoral support and the traditional party with the highest autonomy demandleaving others aside. Finally, while the SAS is able to account for the extent of these parties' autonomy demands, it does not determine how salient the demands are for them. While beyond the scope of the present research, integrating this dimension in demand measurements certainly deserves further scholarly attention.
Despite these caveats, the SAS makes three main contributions to regional studies and comparative politics more generally. First, by accounting for when autonomy demands and statutes evolve from administrative to legislative autonomy, when they evolve in prerogatives and when they evolve in policy competences at the same time, the SAS allows to complement existing accounts of overall decentralization trajectories (e.g., Dardanelli et al., 2018;Lane & Ersson, 1987/1999 by identifying with greater precision the politically relevant moments and anchors in state restructuring processes and to measure the extent of autonomy demands and statutes in a directly comparable and non-distorted way. This should, in turn, also allow to reassess their determinants and consequences with an account of these relevant moments and anchors (e.g., Filippetti & Sacchi, 2016;Keman, 2000;. Second, by being available with the exact same type of scores for both autonomy statutes and demands within sub-state entities, the SAS allows to not only assess the impact of decentralization on general party positions (e.g., Massetti & Schakel, 2016) but to directly compare and disentangle the mutual influence of party demands and statutes. The same holds for the comparison of regionalist and traditional party demands themselves. This allows opening and revisiting research agendas around these mutual influences, such as looking deeper into the paradox of federalism for example, which expects autonomy demands and statutes to push each other to a race to the top in the long run (Erk & Anderson, 2009). Since this is particularly relevant in countries with regionalist pressures and centrifugal decentralization trajectories, the first cross-sectional extension efforts should be directed towards such cases (e.g., Italy or France). Third, and finally, through the substantial time period it covers, even before the rise and evolution of autonomy demands and statutes, the SAS allows for long term and counterfactual studies. With these, the SAS aims at providing new perspectives on measuring, understanding and accounting for modern state restructuring.