Dissecting Public Opinion on Regional Authority: Four Types of Regionalists Based on Citizens’ Preferences for Self-Rule and Shared Rule

Although regional governments play vital roles in most political systems, citizens’ perceptions regarding regional authority are only rarely studied. Relying on the International Constitutional Values Survey held among more than 6,000 respondents from142 regions in eight countries, we develop measures to tap into citizens’ preferences for self-ruleçi.e., for autonomy for their regionçand citizens’ preferences for shared ruleçi.e., for regional engagement in national decision-making. A majority of citizens prefer their regional governments to have some level of both self-rule and shared rule, but around a quarter of the citizens prefer their region to have more self-rule and less shared rule or viceversa. The analysis reveals these varying preferences are associated with a region’s actual authority and regional identity. These results are important because they indicate that most citizens do not presume increased self-rule to be the main or only path to a strong regional authority.

indicators that tap into preferences for a more autonomous sphere for the regional government (self-rule) and preferences for engagement between national and regional governments (shared rule). Exploiting the unique International Constitutional Values Survey (ICVS) held among more than 6,000 respondents from 142 regions in eight countries (Australia, Belgium, Canada, France, Germany, Switzerland, the United Kingdom, and the United States), we reveal that citizens can be divided into four groups based on their attitudes to regionalism. Most respondents (59 percent) can be classified as dual regionalists who prefer their regional governments to have high levels of both self-rule and shared rule, while nonregionalists (16 percent) do not prefer either. Cooperative regionalists (17 percent) prefer shared rule over self-rule, while independent regionalists (9 percent) prefer self-rule over shared rule.
The second contribution is to explore the drivers of the differences in the distribution of different types of regionalists between countries and across regions within countries. An institutional hypothesis expects that citizens' preferences for self-rule and shared rule are associated with the levels of self-rule and shared rule citizens experience in their countries and regions. Higher shares of cooperative regionalists can be found in federations with high levels of shared rule (Australia, Belgium, and Germany), whereas dual regionalists can be found in federations with low levels of shared rule (Canada, Switzerland, and the United States). By contrast, relatively higher shares of independent and nonregionalists can be found in the unitary countries of France and the United Kingdom. An identity hypothesis evaluates whether respondents from regions with strong identities have a higher probability to be an independent regionalist. The results confirm that language regions-i.e., regions where a majority of people speak a different language than the rest of the state-have the highest share of independent regionalists. However, this share declines when a language region is differentiated, i.e., has a unique autonomy arrangement that sets them apart from other regions in the country.
These findings are important because they reveal that preferences for regional government are not a simple zero-sum game whereby citizens simply prefer more or less autonomy for their region. Some citizens prefer more autonomy for their region whereas other citizens would like to see more engagement between regional and national governments. Yet other citizens want their regions to have more or less of both.
The next section develops indicators for citizens' preferences for self-rule and shared rule and categorizes citizens into nonregionalists, independent regionalists, cooperative regionalists, and dual regionalists. We then discuss hypotheses, before explaining the data, variables, and model, and then presenting the results of the empirical analysis.

Citizens' Preferences for Self-Rule and Shared Rule
Regional authority relates to the constitutional arrangement underlying the system of government, given that most countries grant at least some autonomy to local and regional governments through their constitutions. To date, most research into citizen values and preferences with respect to regional authority has been undertaken by federalism scholars, interested in understanding regime support in federations as an enduring, "diffuse" form of support for the overall system of government (Easton 1975;Hobolt and de Vries 2016, 415), as opposed to "specific" support relating to policies and actions of political actors operating within the system (Norris 1999). The authority exercised by regional governments can be usefully described by self-rule and shared rule, two concepts that are widely familiar in the study of federalism. Self-rule is the authority that a regional government exercises in its own territory and shared rule is the authority that a regional government co-exercises in the country as a whole (Elazar 1987;Hooghe et al. 2016, 23;Riker 1964). Recently, Hooghe et al. (2016) have argued and empirically shown that self-rule and shared rule are also useful concepts to explore regional authority in nonfederal countries. Finland, Italy, Portugal, Spain, and the United Kingdom are the examples of European countries that do not have a federal constitution, yet regions in these countries exercise significant shared rule.
Self-rule and shared rule are distinct concepts, but do citizens differentiate between these two concepts when they think about regional authority? Recent research suggests that citizens who favor strong regional authority do not necessarily see this simply in terms of self-rule, and in fact may differentiate between self-rule and shared rule. Surveying multi-level citizenship in fourteen regions in five countries, Henderson (2014) found that a clear majority across all regions in both unitary and federal countries indicated that the region should have greater control over political affairs than the national government. At the same time, citizens who desire greater regional authority also indicated that they desire less regional control in some policy areas and more national uniformity in policy. The results could be contradictory because the logical corollaries of greater regional control and having more regional influence over policy would be more policy diversity across the regions. However, these results could be explained by citizens having different preferences for self-rule and shared rule. Among the citizens who wish greater influence for their region, some think about self-rule-i.e., regional autonomy to decide over policy-whereas others think about the shared rule-i.e., regional engagement and influence in nationwide policy.
The International Constitutional Values Survey (ICVS) enables an exploration into the question of whether citizens differentiate between their preferences for self-rule and shared rule. We draw upon the second wave of the ICVS Brown, Deem, and Kincaid 2018) because it includes three countries (Belgium, France, and Switzerland) in addition to five countries that were part of the first wave (Australia, Canada, Germany, the United Kingdom, and the United States). A stand-alone survey in Australia was fielded online between 3 and 10 August 2017 by OmniPoll through panels managed by LightSpeed Research. The other seven countries were surveyed by KantarTNS, who used the Ncompass internet omnibus survey between 19 and 24 April 2018. To achieve maximum representativeness, sample quotas were set for each region, sex, and age. More detail on the ICVS is provided by  and Brown, Deem, and Kincaid (2018). Table 1 provides an overview of the included regions and the number of respondents for each of the eight countries.
Six survey items from the ICVS, developed as part of a body of research on federal political culture and constitutional values, can also be used to construct indicators for specific preferences for self-rule and shared rule (table 2). Items 1, 2, and 3 tap into preferences for having divided powers between central and regional governments allowing legal diversity across regions as well as a basis to hold each other accountable. These items tap into citizen's preferences for an autonomous sphere for the regional government which connotes the concept of self-rule. Items 4, 5, and 6 measure preferences for engagement and involvement in decisionmaking between regional and central governments. These three items gauge citizen preferences for collaboration between national and regional governments which closely resembles the concept of shared rule. The survey questions and items deliberately did not include the terms "federal," "regional authority," "self-rule," or  (Brown et al. , 2018. "shared rule" to avoid complications arising from a lack of familiarity with these terms among citizens and to allow for comparisons between federal and nonfederal countries Brown, Deem, and Kincaid 2018). The first question is whether these six survey items measure citizen's preferences for self-rule and shared rule. Table 3 displays the results of three principal factor analyses. The first analysis includes all six survey items, the second analysis includes Items 1, 2, and 3-thought to gauge preferences for self-rule-and the third analysis includes Items 4, 5, and 6 which are thought to tap into preferences for shared rule. In each exploratory factor analysis, one latent factor was retained that had an eigenvalue of above 1. Table 3 also displays Cronbach alpha's for each of the three sets of survey items. A Cronbach alpha of above 0.70 is generally considered to indicate a sufficiently reliable measurement. This means that the six survey items reliably tap into preferences for regional authority. The Cronbach alpha's for self-rule and shared rule is a bit lower than 0.70 but the factor loadings Different levels of government being forced to respect each other's roles and responsibilities when dealing with a problem Regional authority 1-6 All six items Notes: The items were preceded with the following question: "Please state if you think each of these is a desirable feature, or an undesirable feature of having different levels of government." The order of the statements was randomized. Respondents had five answer options: very desirable (¼4), somewhat desirable (¼3), somewhat undesirable (¼2), very undesirable (¼1), or cannot say (excluded from the analysis). and the explained variances are higher when the survey items are used separately to assess preferences for self-rule or shared rule. Online table A1 reveals that similar results are obtained when the factor analyses are performed for each country separately. While these results arising from the principal factor analyses may appear somewhat inconclusive, this is in fact explained by the core issue of this research note-that different groups of respondents associate regional authority with selfrule, with shared rule, or with both. Table 4 shows the number and percentage of respondents that combine low or high scores for self-rule, with low or high scores for shared rule. We averaged scores across Items 1-3 (preferences for self-rule) and across Items 4-6 (preferences for shared rule) whereby the averages are re-scaled so that they vary between a minimum of 0-very undesirable across all items-to a maximum of 1-very desirable across all items. The result is to identify that citizens' preferences with regard to regional authority fall across four different groups, which we have labeled as shown. Online Figure A1 displays scatterplots of self-rule and shared rule preference scores by country.
As shown, citizens are not singular in their preference with many preferring different complexions of self-rule and shared rule. Almost three-quarters of the total number of respondents have either high (58.6 percent) or low (15.5 percent) preference scores for both self-rule and shared rule. While this may indicate that around three-quarters of the respondents conceive regional authority as one latent Table 3. Principal factor analysis on six survey items tapping into preferences for regional government Survey item Preference for dimension and evaluate all six survey items accordingly, the fact remains that around one-quarter of the respondents combine a high preference score for selfrule with a low preference score for shared rule (9.3 percent) or vice versa (16.6 percent). In other words, about a quarter of the respondents differentiate between their preferences for self-rule and shared rule, preferring one above the other. Table 5 displays the shares of types of regionalist by country and Online table A2 provides descriptive statistics of self-rule and shared rule preference scores by type of regionalist. Importantly, the four types of regionalists can be found in each of the eight countries but the shares vary considerably. Germany (37 percent) and Switzerland (29 percent) have relatively high shares of cooperative regionalists, whereas many dual regionalists can be found in Canada (70 percent) and in the USA (67 percent). Nonregionalists are more present in Australia (18 percent), France (16 percent), the United Kingdom (18 percent), and especially in Belgium (33 percent).
Future research can and should further explore which citizens conceive regional authority as a one-dimensional issue and which citizens associate regional authority with self-rule or shared rule, including through more sophisticated survey items. However, by way of preliminary investigation, our present data provide further empirical evidence that citizens do differentiate between their preferences for self-rule and shared rule, by revealing how the complexions of citizens' preferences are associated with the authority regime and identity of their region. To provide a clear discussion of the hypotheses, our analyses below compare the four categories of respondents using the labels displayed in table 4. Respondents who prefer their regions neither to have self-rule nor shared rule are named nonregionalists, whereas respondents who prefer their regions to have much of both are labeled dual regionalists. Cooperative regionalists are respondents who Notes: Shown are the number and percentages out of the total number of respondents (n ¼ 6,116) who have scores above or below 0.5 on preference for self-rule and/or shared rule.
prefer shared rule over self-rule and independent regionalists are respondents who prefer self-rule over shared rule. In the next two sections, we develop an institutional and an identity hypothesis to confirm and help explain the distribution of these different types of regionalists.

Impact of the Regional Authority Regime
Several scholars argue that regional authority may produce differences in citizens attitudes across regions. Subnational governments have the capacity to "inculcate in their population distinct political norms," which may lead to differences in attitudes between two regions within a single country "even though they may share demographic or economic characteristics that might otherwise produce similar outlooks" (Henderson 2010, 470). A number of scholars argue that "citizens get what they want," at least in the long term, with regard to regional authority. Livingston (1952, 90) claims that varying degrees of federalism are produced depending on the demands for protection and articulation of diversities, and countries where the demand for integration is stronger than the demand for decentralization will produce (more) unitary institutions. There is some survey evidence that suggests that preferences for federalism covaries with the federal institutional architecture (Cole, Kincaid, and Rodriguez 2004;Scheller 2018). Although there is a debate whether decentralization causes a change in citizens' preferences for regional authority or vice versa, probably most scholars would agree that there is a (strong) association between the two (Erk and Anderson 2009;Jedwab and Kincaid 2018). An association between public opinion on regional authority and the institutional structure of regional authority is taken as a starting point. To the extent that regional authority impacts citizen preferences, one may expect citizens to hold different views about the desirability of self-rule and shared rule. Similarly, to the extent that citizens' preferences impact the allocation of authority between national and regional governments it may do so differently for self-rule and shared rule. The upshot is that the autonomy arrangement of a region co-varies with citizens' preferences for self-rule and shared rule. Citizens from unitary countries may tend to associate regional authority with self-rule because, in contrast to citizens from federations, shared rule is not such a familiar phenomenon. In addition, and also in contrast to citizens from most federations, they are less used to strong regional governments. We expect to find the highest shares of nonregionalists in the unitary countries of France and the United Kingdom. In case citizens from these countries prefer more authority for their regional governments, then they are inclined to prefer their region to have more self-rule than more shared rule and, therefore, we also expect to find the highest share of independent regionalists in these countries. Among the federations in the ICVS dataset-where all regions have high levels of self-rule-we expect higher shares of independent regionalists in federations with low levels of shared rule (Canada, Switzerland, and the USA) whereas we expect to find the highest shares of cooperative regionalists in federations that have high levels of shared rule (Australia, Belgium, and Germany). In comparison with the unitary countries, all federations should have higher shares of dual regionalists.

Impact of Regional Identity
There is overwhelming evidence that citizens with strong regional identity or attachment to their region tend to be also strong supporters of strong regional government (Howe 1998;Liñeira and Cetr a 2015;Llera 1999;Pattie and Johnston 2017;Serrano 2013). Most of this survey research is conducted in the "usual suspect" regions such as the Basque Country, Catalonia, Flanders, and Quebec where regional identities are exceptionally strong and where public support for autonomy and secessionism are comparatively high. However, regional identities also differ widely in countries which are not typically associated with strong national identities such as Australia , 1160-1161, Austria (Erk 2004;Fallend, Ulram, and Zugmeister 2014, 112-115), France (Pasquier 2014, 139-142), and Germany (Oberhofer et al. 2014, 85-87).
Survey research suggests that regional identity may have a greater traction on preferences for self-rule than on preferences for shared rule. For example, Henderson (2014, 168-173) finds that citizens who have stronger regional identities (Catalonia and Scotland) tend to favor more regional control over Public Opinion on Regional Authority political affairs and various policy areas as well as less policy uniformity across the country when compared with citizens from regions where regional identities are weaker (Galicia and Wales). Unfortunately, the ICVS survey did not include a question on citizens' identification or attachment with their region, 1 but our analysis can proceed using a language region dummy variable which scores positive for regions with typically strong regional identities and autonomy demands (Baldacchino and Hepburn 2012; Fitjar 2010) and taps whether citizens in a region speak a different language from that spoken in the rest of the country.
Regions with strong regional identities tend to have a different autonomy arrangement in comparison with other regions in the same country. For example, Scotland and Wales have their own autonomy laws, and Quebec exercises some additional competences in comparison with the other Canadian provinces (e.g., immigration into the province). As mentioned above, most scholars would agree that citizen preferences for regional authority co-vary with the institutional architecture of their region. To "isolate" the impact of regional identity, we include a differentiated region dummy in the models below which scores positive when a region has an authority arrangement that sets it apart from the other (standard) regions within the country.

Method and Data
Our dependent variable is a categorical variable that taps the type of regionalist for each respondent (table 4). Scores for actual regional self-rule and shared rule, as opposed to citizen preferences, are provided by Hooghe et al. (2016). Self-rule is the authority exercised by a regional government over those living in its territory and is measured by a scale of 0-18. Shared rule is the authority exercised by a regional government or its representatives in the country as a whole and is measured on a scale ranging from 0 to 12. 2 Table 6 presents these actual self-rule and shared rule scores for the standard and differentiated regions (in italics) included in the ICVS-dataset. Shair-Rosenfield et al. (2020) classify regions on the basis of whether they have a special authority regime and hence, as explained above, the models below include a differentiated region dummy that scores positive when a region has an authority arrangement that sets it apart from the other (standard) regions within the country (see table 6). We also include a language-region dummy (Shair-Rosenfield et al. 2020), which scores positive when a majority of regional citizens speaks a different language from the dominant national language: Flanders and Wallonia in Belgium; Quebec in Canada; Alsace, Bretagne, and Languedoc-Roussillon in France; Bavaria in Germany; and French-speaking cantons in Switzerland: Fribourg, Geneva, Jura, Neuchatel, Valais, and Vaud.
The models below include a battery of control variables at the individual level which are generally included in models analyzing public opinion on subnational government (Henderson 2014): gender (0 ¼ male; 1 ¼ female), age category (1 ¼ below 30; 2 ¼ 30-39; 3 ¼ 40-49; 4 ¼ 50-59; 5 ¼ 60 and above), and education (0 ¼ less than university education; 1 ¼ university education). In addition, the models include political interest ("How interested would you say you are in politics?"; from 1 [not at all interested] to 4 [very interested]), ideological selfplacement ("Many people use the terms "left" and "right" to describe different political views. Thinking of your own political views, how would you position yourself?"; scores vary from 1 [left] to 11 [right]), and satisfaction with democracy ("On the whole, how well do you think democracy works in [insert country] today?"; from 1 [not at all well] to 4 [very well]). We also include a measure of relative trust in regional government which is derived by subtracting trust in regional government from trust in national government ("Overall, how much trust and confidence do you have in each level of government to do a good job carrying out its responsibilities?"; from 1 [none at all] to 4 [a great deal]). Scores vary from minus 3-no trust in regional government but a great deal of trust in national government)-to plus 3-a great deal of trust in regional government but no trust in national government. 3 Respondents who opted for the "can't say" option are excluded. Descriptive statistics for the dependent and independent variables are provided in Online table A3. We estimate two models, first excluding and then including country dummies. Standard errors are corrected for clustering of respondents by region in both models. We run the first model without country dummies because self-rule and shared rule scores tend to largely vary at the country level (table 6) and the inclusion of country dummies will prevent observing the impact of these variables (Bell and Jones 2015;Plümper and Troeger 2019). However, the share of different types of regionalists varies considerably across the eight countries (table 5), making it important to control for these varying levels through the inclusion of country dummies while estimating the impact of other variables. The full model results are presented in the online appendix (table A4).

Results
First, we present the results for the model that does not include country dummies (Model 1 in Online table A4). Table 7 presents predicted probabilities for the distribution of the four types of regionalists based on the actual self-rule and shared rule scores for the countries/regions, as displayed in table 6. The predicted probabilities were obtained by running margins commands in Stata. The countries/ regions are ordered by their self-rule and shared rule scores. Regions in the United Kingdom and France have limited self-rule and no shared rule; Canada, the USA, and Switzerland are decentralized federations that combine high scores for self-rule with relatively low scores for shared rule; and Belgium, Australia, and Germany are centralized federations where regions have high scores for both self-rule and shared rule. In Online table A5, we present differences between predicted probabilities and their statistical significance.
The results presented in table 7 (and Online table A5) confirm our institutional hypothesis on the regional authority regime, with some exceptions. We hypothesized that all federations should have higher shares of dual regionalist compared with the unitary countries of France and the United Kingdom. We find the expected result for the United Kingdom, but France has similar shares of dual regionalists as the decentralized federations of Canada, Switzerland, and the United States. As expected, the United Kingdom (20 percent) and France (12 percent) have the highest shares of nonregionalists, but France has similar shares of nonregionalists as Australia, Belgium, and Germany (i.e., the differences between the probabilities to find a nonregionalist do not reach statistical significance). The highest shares of cooperative regionalists can indeed be found in the federations that have high levels of shared rule, with the differences in predicted probabilities statistically significant for all comparisons between Australia, Belgium, and Germany (between 22 and 25 percent) on the one hand and the other countries (16 percent or below) on the other hand.
We do not find any impact of self-rule and shared rule scores on the shares of independent regionalists across the countries, but the results for the variable language region supports our regional identity hypothesis. In table 8, we present the results for the differentiated region and language region variables. Table  8 presents predicted probabilities for the different types of regionalists, with results based on Model 2 in Online table A4 including country dummies. Both the language region and differentiated region variables have the most traction on the probability of finding independent and cooperative regionalists. The distribution of the various types of regionalists does not differ between language and differentiated regions and nonlanguage and nondifferentiated regions. The highest share of independent regionalists (18 percent) can be found in language regions that are not differentiated (e.g., the French-speaking Swiss cantons and Alsace and Bretagne in France). The highest share of cooperative regionalists (24 percent) can be found in differentiated regions that are not a language region (e.g., Brussels, Scotland, and Wales). These results confirm our expectation that higher shares of independent regionalists can be found in language regions.
The results in table 8 suggest that a differentiated authority regime may have an appeasing effect on independent regionalists. The results should be interpreted with great care because Flanders, Wallonia, and Quebec are the only three language regions that are also differentiated regions. Nevertheless, the presence of independent regionalists is eight percentage points lower in language regions that have a differentiated authority regime (10 percent) versus language regions without a differentiated authority regime (18 percent). However, we also find that the share of cooperative regionalists is six percentage points higher in the former (21 percent) than in the latter (15 percent), perhaps suggesting that cooperative regionalists may prefer further shared rule reforms.

Discussion and Conclusion
This research note presents empirical evidence that many citizens differentiate between preferences for self-rule and preferences for shared rule. Almost 60 percent of citizens prefer self-rule and shared rule for their region within their national system of government. Around a quarter of citizens prefer their region to either have more self-rule and less shared rule, or less self-rule and more shared rule. Preferences for self-rule and shared rule vary significantly across countries and regions, but do so in predictable ways. Citizens from regions with more self-rule prefer more self-rule whereas citizens from regions with more shared rule have a higher preference for shared rule. Furthermore, the empirical evidence strongly suggests that preferences for self-rule and shared rule have different causes. Citizens from language regions prefer self-rule over shared rule whereas citizens from differentiated regions prefer shared rule over self-rule. Given the novel measures and limited number of countries included in the ICVS dataset, the generalizability of the results is so far limited. It would be interesting to observe citizens' preferences for self-rule and shared rule across more unitary countries and regionalized countries such as Italy and Spain. However, overall, it becomes clearer that citizen preferences regarding regional authority are a result of a combination of preferences for self-rule and preferences for shared rule, rather than simply dominated by demands for and against self-rule alone. This makes it important to differentiate between preferences for self-rule and shared rule to better understand what kind of regional authority citizens prefer. The results shed new light on the often-repeated argument that "federal political culture" and "diffuse regime support" are important for the legitimacy of the constitutional arrangement underlying the system of government and, in particular, of the allocation of authority across national and regional governments, by showing that the way different levels of government do and do not relate may be more important to that culture and support than previously assumed. Further research can and should now also seek to explore differences between preferences for self-rule and shared rule among individuals. While the models in this research note introduced individual-level variables as controls and some statistically significant relationships surfaced (see Online table A4), for now, these have at best tentative explanations (see also Brown 2013). Most surveys ask citizens whether they prefer more or less powers for their regions or whether they are in favor or against further decentralization-without specifying what kind of reform citizens prefer. However, if citizens differentiate between their preferences for self-rule and shared rule, as shown here, we may also expect that these preferences have a different impact on policy preferences. For example, citizens who prefer more self-rule for their region should be more likely to favor more autonomy for their region, less central government interference, and more policy diversity across the regions. Citizens who wish more shared rule for their region can be expected to be more in favor of intergovernmental transfers, more central government involvement, and more policy uniformity across the nation, although presumably informed by rather than wiping out regional perspectives. Similarly, differentiating between citizen preferences for self-rule and shared rule may significantly increase our understanding of when and how decentralization or regional reform satisfies citizen's demands. Far from more or less autonomy for the region being the only options for reform, the evidence presented in this research note. strongly suggests that a substantial proportion of citizens do not think about regional reform as a zero-sum game.

Supplementary Data
2. The RAI taps into formal and legal authority, which may leave out other sources of authority and influence that regions may exercise (Hooghe et al. 2016, 19˚23). 3. Household income is excluded because its inclusion drops the number of respondents by 652. The main findings remain robust when household income is included in the models.