competition over decentralisation: the influence of ideology and electoral incentives on issue emphasis.

Under what conditions is decentralisation a salient issue for state-wide political parties? It is argued in this article that the extent to which state-wide parties emphasise decentralisation depends on their strategic considerations: on their overall ideology, on the electoral incentives created by the context in which they compete, and on the interaction between the two. The results of the analysis of party manifestos in 31 countries since 1945 are as follows. First, parties that pay greater attention to cultural matters relative to economic matters tend to talk more about decentralisation. Second, the systemic salience of decentralisation also encourages parties to talk more about decentralisation. Third, the larger the regionally based ethnic groups within a country, the more salience all state-wide political parties will attach to decentralisation. Finally, only parties that put greater relative emphasis on cultural matters tend to respond to the electoral threat of regionalist parties. The influence of territorial diversity on the salience of decentralisation thus works through two channels and is partly conditioned by political parties’ ideological profile.

have either focused on the influence of party organisation in shaping the incentives for regional actors to challenge the territorial allocation of authority (Riker 1964;Garman et al. 2001;Filipov et al. 2004) or studied how the preferences of actors shape constitutional reforms (Banting and Simeon 1985;Behnke and Benz 2009;Benz and Colino 2011). But we know little about how decentralization becomes an issue in the first place.
To answer this question, we argue that it is necessary to examine the conditions under which state-wide political parties emphasise the issue of decentralisation. To identify those conditions, we follow the encouragement of Marks and Hooghe (2000: 811) to 'bring politics into the study of institutional change'. The premise of this article is that, much like European integration, the territorial structure of the state is a politicised question because it touches policy areas that are close to the hearts and minds of citizens, such as health and education, and because it furnishes regions with a political mantle that potentially challenges citizens' sense of national identity. As a result, we expect that a party's decision to address the state's territorial structure will be sensitive to its ideology. But parties are also responsive to their environment and adjust to the imperatives of party competition. So, we also expect that the emphasis that state-wide parties put on the issue of decentralisation will be shaped by the presence of a territorial cleavage and its politicisation by regionalist parties. Thus, the salience of decentralisation as an issue for state-wide political parties is explained by strategic considerations: political parties talk more about decentralisation if this is consistent with their ideology and with their electoral incentives.
Using manifesto data from the CMP/MARPOR (Budge et al. 2001;Klingemann et al. 2007;Volkens et al. 2012), we first show that state-wide parties that focus on cultural rather than economic issues are more likely to address decentralisation, because this topic is associated with other 'New Politics' issues such as minority rights. But, the context of political competition matters as well. Related to this, we demonstrate that systemic saliencethe extent to which other parties in the party system talk about decentralisationinfluences the party under consideration. We also reveal that a country's degree of territorial diversity matters: where there are large regionally-based ethnic groups, the issue of territorial autonomy is generally more likely to be addressed by state-wide parties. Finally, we show that party ideology and electoral incentives reinforce one another: state-wide parties that focus on cultural issues talk more about decentralisation when faced with the electoral threat of regionalist parties. Thus, the effects of the two main strategic driversideology and electoral incentivesare conditional on one another.
Our approach adds to existing analyses of the salience of decentralisation among statewide parties undertaken by Mazzoleni (2009) and Alonso (2012). 1 These authors seek to understand the 'contagion' of support for decentralisation across national party system, using the CMP as evidence: Mazzoleni's (2009) account indicates that electoral defeats and the electoral threat of regionalist parties play an important role in determining the salience of decentralisation, while Alonso (2012) identifies a conditioning effect of ideology in shaping the ability of mainstream parties to adopt a credible pro-devolution strategy. We extend these findings in two ways. First, we consider how the importance of decentralisation is conditioned by the emphasis placed on other issue dimensions (economic, cultural) which constitute the 'package' of party policies. Second, we look at how state-wide parties respond to strategic incentives across a wide universe of cases. Mazzoleni (2009) and Alonso (2012) apply their arguments to a select number of West European countries (Belgium, France, Italy, Spain and the UK), where the territorial cleavage and decentralisation processes have been prominent. 1 While there have been efforts to study state-wide parties' positioning on decentralisation (e.g. Toubeau and Wagner 2013), we note that studying party positioning on the issue needs to be distinct from the analysis of salience: the position a party takes on an issue is not necessarily related to the emphasis it places on it, even if there is evidence that some parties emphasize their more extreme positions (Wagner 2012).
But this means that the broader applicability of their explanation is limited and that we know little about why decentralisation becomes salient in different kinds of settings, i.e. homogeneous or heterogeneous countries, and, in the latter case, whether it is driven by the threat of regionalist parties or by territorial diversity tout court.
The next section develops the theoretical reasoning underlying our claims. In the third section we describe the data, measures and statistical models that we employ to assess these claims. The fourth section describes the results. The conclusion summarises the findings and discusses their relevance to the study of issue competition and multi-level governance.

When do parties emphasise decentralisation?
Recent decades have seen a rise in issue voting and competition (Franklin et al. 1992;Green-Pedersen 2007), in which topics such as the environment, immigration and the EU have become increasingly influential in shaping voter choice and party strategy. From the perspective of political parties, decentralisation is one among this set of policy issues over which they compete for electoral support. We argue that, as with other issues, the emphasis that political parties put on decentralisation is determined by strategic considerations: specifically, by their ideology and by their electoral incentives.
Our first contention is that state-wide parties' policy stances on decentralisation are part of their overall ideology, i.e. the set of values, goals and beliefs about societal institutions that define their identity and guide their actions (Freeden 1998). We posit that the ideology of parties encourages them to prioritize certain issues above others and that, during electoral contests, they place a selective emphasis on those higher priority issues because it confers to them a strategic advantage (Budge and Farlie 1983). Therefore, the salience of decentralisation is likely to be conditioned by the emphasis placed on the other issues that constitute their programmatic 'package', which can be organised along the economic and cultural dimensions (e.g., Kriesi et al. 2008).
Specifically, parties that put greater emphasis on the cultural dimension relative to the economic dimension should be more likely to talk about decentralisation. We expect this because decentralisation has come to be associated with the post-materialistic values that grew among Western electorates in the late 1960s and early 1970s, and with the subsequent rise of Green parties and 'New Politics' (Inglehart 1990). Decentralisation was seen as a method for fostering greater participation in decision-making and granting collective rights to autonomy for particular territorial groups, and thus became prominent alongside other 'new' issues related to the cultural dimension, such as the environment and the collective rights of non-economic groups like women, ethnic minorities or immigrants (Marks et al. 2010). In response to this evolution, there occurred a parallel rise of Radical Right parties during the 1980s that emphasised the flip-side of these post-materialist issues (Mudde 2007). These culturally conservative parties adopted radically opposite stances, favouring traditional morality, nationalism and central state authority.
State-wide parties from mainstream families such as Social Democrats and Liberals also articulated the cultural dimension as a result of its prominence in public opinion and the electoral threat of the new 'niche' parties (Meguid 2008). Parties that respond in this way should also be likely to talk more about decentralisation. In contrast, mainstream parties which maintain their focus on the economic dimension should not emphasise decentralisation as much. Of course, all political parties will devote some attention to both the cultural and economic dimension. However, the relative emphasis on the two dimensions will vary, depending for instance on the extent to which mainstream parties have incorporated 'New Politics' issues. This should have a direct bearing on the importance they assign to decentralisation. So, our first hypothesis is: H1: The more a state-wide party emphasises the cultural dimension relative to the economic dimension, the more it will emphasise decentralisation.
Our second contention is that parties adjust the salience of decentralisation in their programmes in accordance with the electoral incentives shaped by their environment.
Specifically, we argue that parties will alter the emphasis they put on decentralisation in response to the territorial diversity of a society and its politicisation by regionalist parties.
First, state-wide parties should place greater emphasis on decentralisation in countries with regionally-based ethnic groups, such as the Scots in the UK or the Flemish in Belgium.
We define a regionally-based ethnic group as a group of people living in a territorially delimited space that shares a sense of commonality based on a belief in a shared ancestry and a common culture, that is politically relevant insofar as it is represented in national politics by a least one political organisation (Cederman and Girardin 2007;Cederman et al. 2010). It is in those contexts that decentralisation is politicised because it has ambiguous consequences for the mobilisation of ethnic grievances. 2 The size of the regionally-based ethnic group is a key determinant underlying both their predisposition to advance secessionist claims (Sorens 2005(Sorens , 2008 and the willingness of states to accommodate their demands (Walter 2006). This generates the expectation that political parties will pay greater attention to decentralisation if regionally-based ethnic groups are large. This is because it is a potential threat to the integrity of the state and thus an unavoidable issue of debate, and because there may be a strategic incentive to check the future actual threat of regionalist parties.
Once this actual threat is present, so when there are relevant regionalist parties, statewide parties may have an even stronger incentive to talk about decentralisation. By regionalist parties, we denote parties that represent territorially-bounded ethnic, linguistic, cultural groups, that seek electoral support on a limited territorial basis and whose main goal is selfdetermination, i.e. the right to exert direct control over their ruler and policies, whether in the form of territorial autonomy or independence. In virtue of their presence and demands, regionalist parties can introduce decentralisation as a separate issue dimension of competition over which they can claim ownership as niche political parties (Meguid 2008;Wagner 2011) and through which they can exert pressure on state-wide parties in the electoral, parliamentary and governmental arenas (Toubeau 2011;Amat and Falco-Gimeno 2013). Following Meguid (2008), we expect that if the size of regionalist parties is small and the actual threat therefore negligible, on the whole state-wide parties will likely dismiss the issue as unimportant.
Conversely, if the size of regionalist parties is large and the actual threat therefore significant, we expect state-wide parties in general to increase the priority they assign to decentralisation. They do this either by deploying an accommodative strategy in an attempt to challenge regionalist parties' ownership of the issue or by deploying an adversarial strategy in order to under-cut the accommodative efforts of their mainstream rivals (Meguid 2008). Our two hypotheses relating to potential and actual threats created by territorial diversity are thus: H2a: The larger the size of the regionally-based ethnic groups, the more state-wide parties will emphasise decentralisation.
H2b: The larger the size of regionalist parties, the more state-wide parties will emphasise decentralisation.
Our third claim is that the components underlying a party's strategic considerations (ideology and electoral incentives) reinforce one another. Thus, how parties react to territorial diversity will depend on the configuration of their ideology. The first expectation is that parties that emphasise the cultural dimension are more likely to respond to territorial diversity by emphasising decentralisation than parties that emphasise the economic dimension. Given the importance they assign to cultural pluralism and localised decision-making or, conversely, to centralism and state authority, these parties have the requisite ideological background to address the issue of decentralisation in their strategic response to the threats presented by regionally-based ethnic groups and by regionalist parties. That is, they can expect that putting selective emphasis on decentralisation and thereby raising its prominence in the political space will be favourable to them. The related expectation is that the issue of decentralisation will be more closely associated with the cultural dimension in countries that feature greater territorial diversity. That is because in such contexts, parties that assign importance to the cultural dimension will include decentralisation in their 'package' of cultural concerns. The consequence is that the relative emphasis on the cultural dimension is more closely tied to the salience of decentralisation in countries where the potential or actual threat of territorial diversity is higher. This generates a third set of hypotheses: H3a: The larger the size of regionally-based ethnic groups, the more state-wide parties that emphasise the cultural dimension will emphasise decentralisation.
H3b: The larger the size of regionalist parties, the more state-wide parties that emphasise the cultural dimension will emphasise decentralisation.

Data and Model
To measure the salience of decentralization for political parties, we make use of the party manifestos coded by the CMP/MARPOR (Budge et al. 2001;Klingemann et al. 2007;Volkens et al. 2012). This project summarizes party manifestos quantitatively by assigning each quasi-sentence to one of 56 categories. This approach is useful for our purposes as it explicitly measures the emphasis each party places on the different issues. 3 Following Alonso (2012), we code decentralisation emphasis using six issue categories: decentralisation, centralisation, national way of life (positive/negative) and multiculturalism (positive/negative). These additional categories are not exclusively related to decentralisation, as they may refer to topics related to immigration or integration. But, adding them nevertheless provides us with a more valid measure of the total salience of decentralisation, as it offers a measure of a party's stance towards the institutional and cultural component of decentralisation. 4 Moreover, parties do not always phrase support for the national state and the central government as support for 'centralisation', so such quasi-sentences hardly exist at all.
While robustness checks reveal that our results are not sensitive to the precise categories we include, we follow Alonso's coding as this appears to best reflect actual party salience. 5 Figure 1 presents descriptive graphs depicting the average salience of decentralisation across time, party families and countries. The growing importance of decentralisation is evident from the trend line in the left panel: the attention that parties have paid to this issue was mostly constant from the 1950s and then increased in the 1970s, a period that corresponds to the surge of political nationalism among stateless nations and to the increasing appeal of decentralisation. Yet, there is also a large amount of variation in the salience of 3 We exclude programmes classed as estimates by the CMP itself. 4 Certain scholars have sought to improve the validity of decentralisation scores in the CMP dataset by coding these two different dimensions, but these efforts remain limited to specific countries, like Spain (Libbrecht and Maddens 2009) and Italy (Basile 2012). 5 We also ran our models with just the two decentralization codes as well as adding just the items related to 'national way of life'. The results remain largely consistent (see Appendix 5, Tables 2 and 3), with the partial exception of H3b (Model 4).
decentralisation. This fact is evident in the considerable spread of salience scores around the trend line in the left panel of Figure 1. Moreover, while countries and families differ on average in the importance that parties assign to decentralisation, there are also notable differences within all party families (central panel) as well as within countries (right panel).
The overall picture is thus one of considerable variation in the salience of decentralisation within party families and within countries, and it is this variation we aim to explain.
[ Figure 1] The countries and elections included in the analysis are listed in Appendix 2. We use an OLS model to predict the salience of decentralisation. Since errors may be correlated within parties, we cluster standard errors by party. To address the autocorrelation of errors from one election to another, we run our models using a Prais-Winsten transformation (as recommended by Plumper et al. 2005). We choose this method over the use of a lagged dependent variable as the latter approach arguably uses lagged values to explain much of the variance of interest. 6 To address the possibility that errors within one country-election may be correlated, we use a series of country-and election-level covariates, detailed below. We exclude all parties coded as regionalist (see Appendix 3) from our analysis. In our regression models, we use the natural logarithm of the salience score as is recommended for skewed data that is zero-censored (Gelman and Hill 2007). 7

Predictor variables
To measure our first predictor variable, we create a relative salience indicator for the cultural dimension relative to the economic dimension. We use the coding approach suggested by and Hobolt (2012) (see Appendix 1). 8 The cultural dimension includes topics such as freedom and human rights; traditional morality; law and order; and environmental protection. The economic dimension contains familiar issues such as protectionism, regulation and free markets. We measure the relative salience of the cultural dimension as the share of cultural statements among all economic and cultural statements: salienceculture/ (salienceeconomy+salienceculture). This variable ranges from 0 to 1.
We measure the size of regionally-based ethnic groups using information provided in the Ethnic Power Relations (EPR) dataset (Wimmer et al. 2009, Wucherpfennig et al. 2011 retrieve the proportion of the population belonging to a regionally-based ethnic group. We code ethno-politically relevant groups as regionally concentrated if they are recorded as either only or partly regionally-based in the EPR dataset. We exclude the largest ethno-politically relevant group, which is generally the dominant group (e.g. the English in the United Kingdom). The resulting variable ranges from 0 to 1; the maximum value in the dataset is for Belgium (0.4). Finally, we also code the electoral strength of regionalist parties. To do so, we first created a list of all regionalist parties (see Appendix 2) and then coded their electoral success (using Massetti and Schakel 2013). We use the values from the current (and not the previous) election as we believe that this best captures the electoral threat.

Controls
At the party level, we include two variables that may affect the emphasis that parties place on decentralization. Party size is measured as the share of the vote at the election after the manifesto was written; this information is included in the CMP dataset. Smaller parties should emphasise decentralisation more as they may wish to occupy a niche area in the policy space (Wagner 2011). Government participation is 1 if the party was in government for any amount of time (excluding caretaker cabinets) between the previous and the current election; this information is taken from the Parlgov dataset (Döring and Manow 2012), with missing cases added manually by the authors. Parties in opposition should emphasize decentralisation more than parties in government, who will seek to preserve the institutional status quo.
At the country level, we include eight variables that may shape the strategic incentives faced by political parties to emphasize decentralisation. We control for how other mainstream parties campaign on decentralisation matters by including lagged measures of the systemic salience and party polarisation on decentralisation at the previous election. 9 Polarisation is measured as the standard deviation in party positions at the previous election. 10 We also 9 Appendix 5, Table 4 presents the results if we use concurrent levels of systemic salience and polarization. The results remain substantively very similar; however, H2a receives somewhat less support in this specification, which is not surprising as the effect of regionally-based ethnic groups will affect all parties and thus be partly contained within the systemic salience measure if it is measured concurrently. include the level of disproportionality and the effective number of electoral parties; both variables are taken from Gallagher (2012). A permissive electoral system is likely to facilitate the articulation of the territorial cleavage, the multiplication of issue-dimensions (Anorim Neto and Cox 1997), and to generate a centrifugal dynamic of party competition in which parties seek marginal votes by focusing on non-economic issues like decentralisation (Cox 1990;Dow 2001). Finally, we control for the existing territorial distribution of authority, measured by the 'self-rule' value assigned by Marks et al. (2010). Self-rule is the extent to which sub-national units can run their own affairs independently of the central government.
We use the value from the previous election. The more powers territorial entities exercise, the more likely it should be that decentralisation will be an issue of contestation. The process of reforming the territorial distribution of authority can, however, also be contentious. So, we code whether there was a territorial reform between the previous and the current election: this variable is 1 if there was a change, 0 if not. Alonso (2012) suggests that the salience of centreperiphery matters increases immediately after (and not before) a territorial reform. Lastly, we also control for the population size and geographic area of the country. These are taken from

Results
To test our first hypothesis (H1) that mainstream parties that focus less on the economic dimension and more on the cultural dimension also put greater emphasis on the decentralisation dimension, we run a basic model ( Labour party did during its electoral campaigns in the late 1990s. In contrast, parties that remain concerned with questions of equity, state intervention and economic groups and thus compete mostly on the economic dimension will downplay the issue of decentralisation. Turning to the control variables, it appears to be difficult to explain the salience of decentralisation matters using country-or party-level factors. However, the result for systemic salience is consistently strong and statistically significant: for every 1 percent increase in systemic salience, the emphasis on decentralisation increases by .21 per cent. This means that a 25 percent increase in systemic salience (e.g., from 0.4 to 0.5) would lead to a 5 per cent increase in the emphasis on decentralisation. This reveals that parties do not autonomously determine the importance they attach to decentralisation according to their ideological profile, but rather are responsive to their political environment and strategically adjust their emphasis in function of how much other political parties talk about this issue. The influence of the 'party system agenda' may encourage increased attention to decentralisation even on the behalf of those individual parties that do not 'own' the issue and do not stand to reap electoral benefit from addressing it (Green-Pedersen 2007). This finding is comparable to studies that examine the determinants of the salience of issues like European integration (Steenbergen and Scott 2004). and the environment (Spoon, Hobolt et al. 2014) that show that parties are constrained by their strategic context in deciding the emphasis they place on the issue.
The effect of electoral incentives is also evident when we assess the influence a Why do state-wide parties respond to the potential threat of territorial diversity, but not to the actual threat posed by regionalist parties? The first thing to note is that, in countries with greater territorial diversity, the issue of decentralisation is more politicised at the systemic level, that is, across state-wide parties in general. 13 This is because it is in those settings where decentralisation is deployed for the contentious purpose of managing a territorial cleavage, which has bearing for the entire polity and thus demands to be addressed by all state-wide parties. Moreover, even if this territorial diversity does not find political articulation through regionalist parties, all state-wide parties face the electoral incentive to check the possibility that a potential threat becomes an actual threat. This incentive will naturally rise with the size of the group. This does not mean that the claims of smaller groups, such as the German-speaking minority of South Tyrol, will not be recognized, but rather that it will not feature as saliently as an issue in national elections. This result corroborates existing studies (Sorens 2005, Walter 2006).
But why is there no such effect for the influence of regionalist parties? The answer is that different state-wide parties will face different constraints in their ability to respond to this actual threat, depending on their ideological profile: some parties will respond by talking more about decentralisation, while others will not, thus reducing the effect that regionalist parties may have on the salience of decentralisation, when all parties are examined.
We show this when testing Hypotheses 3a and 3b, in which we argued that some parties may react more than others to the potential and actual threat of territorial diversity, depending on their ideological profile. Models 3 and 4 test whether the effects of the size of regionally-based ethnic groups (potential threat) and the electoral strength of regionalist parties (actual threat) are conditioned by the relative emphasis that a party places on the cultural dimension.
Model 3 tests the effect of the interaction between the relative salience of the cultural dimension and the size of the regionally-based ethnic group on the salience of decentralisation. The interaction effect is not statistically significant. 14 Thus, there is no evidence supporting the claim that the effect of regionally-based ethnic groups on the salience of decentralisation is stronger among parties that put greater emphasis on the cultural dimension, or that the salience of decentralisation is more closely associated with the salience of the cultural dimension in countries with larger regionally-based ethnic groups. The reason for this echoes what we argued above: all state-wide parties face the incentive to address the issue of decentralisation when responding to the presence and size of regionally-based ethnic groups, in order to check the potential emergence of an actual threat, so we do not observe any cross-party variation by ideological type.
In contrast, we find that the effect of the threat of regionalist parties is stronger among parties that put greater emphasis on the cultural dimension (Model 4). To verify the hypothesis, we restrict our sample to countries where there is a regionally-based ethnic group and we interact the electoral strength of regionalist parties with the relative salience of the cultural dimension. The interaction effect is statistically significant, providing support for H3b. In Figure 2, we see that the marginal effect of the electoral strength of regionalist parties on the salience of decentralisation increases as the relative emphasis on cultural matters increases. 15 This means that the effect of the strength of regionalist parties on the salience of decentralisation is conditional upon the fact that a party emphasises cultural matters; in contrast, the influence of regionalist parties is rendered neutral for parties that compete exclusively on the economic dimension. So, the main mechanism or 'transmission belt' from the actual threat of territorial diversity to the heightened salience of decentralisation is the relative emphasis that state-wide parties place on cultural issues.
[ Figure 2 about here] For an illustration of how these mechanisms worked in reality, consider the response of state-wide political parties to regional nationalism in the UK and Spain. During the 1980s, Scottish and Catalan regionalist parties put forth claims for independence and territorial autonomy, but at the time, the British Labour party and the Spanish Socialist Party were primarily concerned with economic matters like trade unions, state ownership of the economy, employment and welfare, and therefore did not much discuss devolution or the empowerment of Autonomous Communities. However, when these state-wide parties relaxed this focus on economic issues in the early 1990s and began to be more concerned with issue institutional reform, group rights and gender issues, then they responded to the claims of nationalist parties and decentralisation became a more prominent topic in their programmes. We first demonstrated that parties that assign greater relative importance to the cultural dimension also tend to give decentralisation a greater role in their party programme. So, parties that talk about safeguarding traditional morals, combatting crime or protecting minorities, and that stay clear of issues relating to equality, state intervention and redistribution, also tend to pay more attention to decentralisation. Second, decentralisation is also more salient among parties that compete in a context characterised by territorial diversity.
More specifically, we found that the two components of territorial diversity work in two distinct ways. The potential threat represented by regionally-based ethnic groups tends to raise the salience of decentralisation among all state-wide parties, irrespective of their ideological profile. In contrast, the actual threat of regionalist parties raises the salience of decentralisation only among political parties that put greater relative emphasis on cultural matters. Thus, the presence of regionally-based ethnic groupthe main characteristic distinguishing homogeneous from heterogeneous societieswill produce an increase in the salience of decentralisation at the level of the party system. Parties are clearly sensitive to their strategic context when they decide the issues on which to focus. This finding is strengthened further when we consider this result in conjunction with the significant effect of systemic salience: parties increase their emphasis on decentralisation if other parties talk about it as well. However, we also found that in heterogeneous countries, a party's responsiveness to the electoral threat of regionalist parties is conditional upon its ideology: only parties which emphasise the cultural dimension will pay greater attention to decentralisation in response to strong regionalist parties.
These findings advance the state of our knowledge on the topicone strongly shaped by the recent contribution of Alonso (2012) -by showing that a party's overall ideology, in particular the relative emphasis placed on the economic and cultural dimensions, shapes the importance that it assigns to decentralisation. By assessing our claims in a broader empirical universe of cases that includes homogeneous and heterogeneous countries, we are also able to show that decentralisation becomes prominent as territorial diversity increases, so as both regionally-based ethnic groups and regionalist parties become larger, but that parties' response to this diversity is conditioned, in part, by their ideological profile. of power in such systems cannot be limited to the narrow question of efficiency, but rather is subject to political contestation that will vary with the ideology of political parties and the structure of their electoral incentives, as shaped by nature of party competition and the articulation of the different identities of territorial groups living in a country. Note: * p<0.05, ** p<0.01, *** p<0.001; standard errors (clustered by party) in parentheses. See supplemental materials for robustness checks using alternative model specifications.