Economic Shocks, Mobilization, and Regional Elite Splits

What are the origins of elite splits? Why do regional elites break away from central elites and develop regional parties? This paper contends that intra-elite differences are more likely to be politicized when an economic shock exacerbates pre-existing asymmetric economic preferences and disadvantaged elites can mobilize the electorate on the basis of identity. I employ constituency-level data from Catalonia spanning the late 19th and early 20th century to test which factors influenced regional elite decisions to form a regional elite political party. To understand elite divisions, I exploit a historical exogenous trade shock and its asymmetric impact within Catalonia, and the availability of identity-based mobilization agents. The results show that regional elite splits took place in areas more affected by the 1898 colonial trade shock and where elites had larger mobilization capacity.


Introduction
When do elites remain united and which processes and preconditions lead to elite splits? Why do some regional elites remain loyal to central peers while other regional elites prefer confrontation through the creation of a new political party? Classical accounts have focused on modernization processes and the existence of cultural or linguistic characteristics to explain political divisions (Lipset & Rokkan, 1967). From this perspective, political (and elite) divisions have usually been taken for granted as an inescapable consequence of social and economic preconditions. However, the cleavage theory is not able to explain when and why certain elites decide to support an elite split.
In this paper, I propose a theoretical framework describing under which conditions regional elites break away from central elites by creating a new regional political party. As such, I answer a classical Rokkanian question (Rokkan, 1970) employing a novel historical trade shock approach that allows to test the mechanisms driving elite splits. Through this approach, I show the micro-level mechanisms behind the political division of elites. 1 Elite divisions and competition are essential to understand the configuration of political institutions and party systems, and divisions generate longlasting consequences (e.g., Ansell & Samuels, 2014;Beramendi et al., 2019). Hence, it is important to dig deeper to better understand the conditions under which certain regional elites choose to break away from central elites. This paper contributes to the literature by clarifying the factors that lead to elite divisions. My main claim is that elites split along (and politicize) a political cleavage when they suffer economic grievances, and they have the capacity to mobilize likely supporters. If both conditions are met, we can expect an elite division to take place. Grievances can foster popular demand for an autonomous political party, and through appealing to these grievances party elites who control mobilization tools can maximize electoral support. Thus, mobilization capabilities are not only relevant to understand democratization (Ziblatt, 2017), but also to understand the origins of elite divisions.
This article focuses on the origins of elite splits along the center-periphery conflict dimension. The rise of center-periphery cleavages has been thoroughly explored by Rokkan and Urwin (1982), who considered their emergence a consequence, on the one hand, of divergent economic development and, on the other hand, of distinct identities. It is undeniable that economic geography and economic asymmetries lead to political differences and territorial tensions (Rickard, 2020) and drive decentralization demands and even secessionism (Sorens, 2012). The materialization of centerperiphery elite splits does not automatically occur in contexts with diverging economic conditions or identities, but are related to grievances (Mor, 2022). Regional elites in territories that would later herald their difference, usually stood together with central elites during initial steps of economic divergence.
To better understand the timing of political decisions, recent literature has focused on trade shocks as plausibly exogenous factors influencing political dynamics (Colantone & Stanig, 2018). A sudden-and geographically uneven-change in economic conditions produces different patterns of political behavior among voters, but also among elites, and it can pave the way for the rise of new political actors (Rodrik, 2018). However, without the capacity to mobilize large segments of the electorate, economic grievances might not suffice to explain the emergence of a center-periphery cleavage. Political mobilization is a key factor that determines the rise of new parties (Kalyvas, 1996;Boix, 2009) and political entrepreneurs ponder all available possibilities before engaging in party formation (Tavits, 2008).
I test under what conditions center-periphery elite divisions materialize focusing on the case of Catalonia. During the 19th century, Catalonia was the most industrialized region in a mostly rural Spain (Nadal et al., 1988), and it was, in addition, a region with a distinct language and a past record of selfgoverning institutions. The preconditions for the emergence of a centerperiphery cleavage were already in place in the 19th century, why then did Catalan elites remain politically aligned with Spanish elites until the early 20th century?
Catalonia is a suitable case to explore the origins of elite divisions because of its internal heterogeneity in the extent and the types of industrialization (Carreras, 2019). The leading textile sector grew extensively during the 19th century and the availability of a Spanish colonial market contributed decisively to this development. The loss of the oversea colonial territories in 1898 was a traumatic critical juncture that compelled Spanish elites to rethink the international role of Spain as well as its political institutions (Álvarez Junco, 2002). Moreover, the colonial defeat had large economic repercussions (Maluquer de Motes, 1999) and it represented an important economic shock to the Catalan economy and its industrial elites. The colonial shock was not only exogenous, but its impact was also heterogeneous across regions and industrial sectors across Catalonia (Harrison, 1974). Moreover, while Catalan identity was being revived by an intellectual intelligentsia in some areas, this revival effort was less prevalent in other areas.
In this paper, I employ a novel dataset that combines data on electoral returns, industrialization, exports, mobilization tools, and other socioeconomic indicators. My dependent variable captures whether regional elite candidates stood for office at the constituency level. To account for economic grievances, I measure constituencies' exposure to the colonial trade shock using an indicator similar to the shift-share instrument developed by Autor et al. (2013). I proxy political mobilization capabilities of regional elites through the geo-referenced origins of delegates attending an important pro-Catalan assembly that took place in 1892.
I argue that the estimates of my analyses can be interpreted causally because the effects of the colonial shock were unexpected and exogenous to the preferences of Catalan industrial elites. I also provide evidence that colonial shock exposure was uncorrelated to previous political outcomes. The results I obtain support the idea that regional elites decided to split from central elites when economic grievances were deeper, and elites could resort to identity-based mobilization. I provide evidence on plausible mechanisms and I show that results were stronger in export-oriented constituencies (e.g., those with a large presence of cotton textile), for years shortly after the colonial defeat, and in legislative (first order) elections.
Although the analyses focus on the Catalan case, other regions, such as Ireland or Bavaria in the 19th century, or the Basque Country in the early 20th century, also saw the emergence of powerful regional elites. Despite the existence of economic and identity preconditions, the differences in timing of these regional elite splits cannot be explained by extant theory. The framework and explanation exposed here thus contribute by shedding new light on when regional elites emerge, beyond the case of Catalonia.
The rest of the paper proceeds as follows. In the next section, I conceptualize elite splits, and I dialogue with the main literature regarding centerperiphery tensions, elite fragmentation, and the impact of trade shocks; I conclude the section by presenting the main expectations to be tested. The third section describes the main characteristics and key historical events of 19th and early 20th century Catalonia. After that, I describe the data set and the operationalization of the main variables, and I detail the empirical strategy. In the following section, I present the main results, followed by complementary analyses that show the specific mechanisms and the robustness of the findings. I end up by summarizing the results and discussing its implications.

Theoretical Framework
In this section, I first define the concept of elites and elite splits, I describe the different available strategies for elites, and I show the relevance of intra-elite splits for different sorts of political outcomes. Subsequently, I overview previous literature on the origins of political cleavages and on elite heterogeneity. Finally, I outline the main hypotheses.

Elites and Elite Conflict
Definition and Strategies. I broadly define elites as a small group of individuals in society who have disproportionate access to political offices, resources, influence, or power (Mosca, 1939). Hence, I consider elites to be economically well-off individuals-for example, industrialists or owners of large companies-individuals with large political influence-for example, MPs or high-level bureaucrats-or individuals with social influence-for example, intellectuals or aristocrats-to name just a few. The canonical literature on elites has highlighted their unitary nature, even though elites can be heterogeneous in the field in which they are powerful (Mills, 1956;Putnam, 1976). 2 However, elites do not always remain united, and their divisions can lead to relevant social and political changes. Therefore, I focus on elite splits-elite divisions or elite fragmentation are employed as synonyms.
I conceptualize an elite split as the development of a new organization involving a subset of elites. An elite split requires two elements: (a) the preexistence of formal elite organizations, (b) the shift of individuals from these pre-existing groups into the new organization. Examples of elite splits could be the creation of a new political party by elites belonging to previous elite-parties or the creation of a new employers' organization by members of existing employers' organizations. This definition does not imply that all members of the new elite organization had to be related to existing elite groups-the new organization can be created to attract different elite groupsbut at least part of the members should shift allegiances in order to consider it an elite split.
The previous definition of elite divisions highlights how elite struggles do not always entail the formation of new organizations; before engaging in elite splits, dissenting elites have multiple available strategies. First, they can remain silent and conform to the existing elite groupings, probably in exchange for perks such as political offices. Second, discontent elites can choose to remain in existing organizations if differences can be accommodated in the existing institutional set up through the development of factions (Kam, 2009). On the one hand, this can be a strategy for existing organizations to attract new emerging elites (e.g., Fresh, in preparation). On the other hand, existing elites who are increasingly discontent can choose to remain in an existing organization and create a faction if either the benefits of breaking the elite agreement are too low or the costs are too high (Weingast, 1997). Finally, if internal factions are insufficient and the benefits of an elite split are higher than the status quo, we should expect the discontent elites to formalize elite divisions (Tavits, 2008), for instance through the creation of a new political party. Although the existence of factions would also reflect divisions or fragmentation within elites, this paper is only interested in formal elite divisions.
Consequences of Intra-Elite Struggles. Formal elite divisions have had important consequences throughout history. For instance, the rise of industrial eliteseconomically powerful but initially politically weak-strained the autocratic nature of many 19th century Western polities. Among many others, intra-elite conflicts have been employed to explain the transitions to democracy (Ansell & Samuels, 2014) or franchise extensions (Lizzeri & Persico, 2004) that eroded plutocratic rule. Moreover, different types of elite-coalitions had longlasting consequences for democratic transition and consolidation (Luebbert, 1991). A growing stream of literature has also recently delved into the consequences of intra-elite conflicts for institutional developments (Paniagua & Vogler, 2022), increases in state capacity (Garfias, 2018), or changes in fiscal institutions and taxes (Beramendi et al., 2019;Mares & Queralt, 2020).
The previous findings show that intra-elite competition is a powerful engine of institutional change. Nonetheless, a shortcoming of this literature is that it always considers intra-elite divisions as given. Although it is plausible to think that conflicts among elites were related to institutional and economic preconditions, it is necessary to endogenize the relationship between the political and economic consequences of elite fragmentation. Elite divisions do not develop equally across countries, and the manifestation of elite splits might also be strategic and conditional upon elites' available resources.
For intra-elite clashes to have political consequences, however, elites that seek political influence need first to gain access to institutionalized political power. Discontent elites need to be formally grouped in new organizations, such as political parties, to be decisive in decision-making processes. Hence, it is important to understand why and when intra-elite divisions are formalized.

Elite Heterogeneity
An initial account to understand political and elite heterogeneity relies on the influential cleavage literature introduced by Lipset and Rokkan (1967). In their work, societies have more-or-less fixed characteristics that determine the social lines of conflict, which have the potential to produce political and elite divisions. Social and individual characteristics define the political preferences and the behavior of actors and, although changes are possible, new cleavages can only emerge in the long-term after processes of profound social transformation.
Similarly, the economic geography literature considers heterogeneity in political preferences to be influenced by differences in economic geography endowments. Given that different territories are subject to distinct resources, structures and dynamics, and sociodemographic compositions, political actors should have different preferences and behaviors (Rickard, 2020). Hence, based on the cleavage and the economic geography literature, intrinsic and almost-fixed factors seem to determine the preferences and the behavior of elites.
A missing aspect in previous studies is the lack of information on when elites decide to break away. According to Weingast (1997), elite unity is generally based on implicit pacts and disadvantaged elites can only break the pacts when the status quo changes, for instance when there are technological changes (Boix, 2015) or sudden economic shocks. Rokkan and Urwin (1982) posited that regional or peripheral identities were politicized in times of industrialization because of geographically uneven economic development. 3 Hroch (1985), in turn, emphasized the role of identity awareness and signaled the crucial role played by "national activists" (p.13) to understand the revival of national movements.
However, less is known about the conditions under which differences in economic preferences or identity lead to party divides. Rokkan & Urwin (1982) defended that modernization, when coupled with a different identity, can generate the division of elites and the emergence of center-periphery cleavages. Unfortunately, the previous literature did not test when the overlap between economic grievances and distinct identity boosts the emergence of center-periphery tensions. 4 Uneven economic development and distinct identity are not sufficient, as these factors were already in place in many territories-such as Catalonia, Scotland, or Corsica-much before than center-periphery tensions were politicized by elites. I argue that, beyond the factors considered before, disruptive economic shocks and mobilization capabilities are fundamental to understand the origins of elite splits.

Expectations: Economic Shocks and Political Mobilization
A sudden change in international trade can entail a significant economic shock with large political effects. For instance, actors exposed to international trade demand compensation for the negative consequences of trade disruptions, usually through larger public investments (Rodrik, 1998). Moreover, recent contributions have shown that trade exposure is related to changes in party systems (Calca & Gross, 2019) and the ascent of new populist political actors (Rodrik, 2018).
To assess the effects of trade shocks, a novel identification strategy based on the use of shift-share instruments has allowed to explore the political effects of these shocks (Autor et al., 2013). These instruments permit to infer the exposure to trade at micro-level units despite the lack of disaggregated trade data. Different papers using these instruments have shown that exposure to import trade shocks has had important effects on individual-level political preferences and behaviors (Colantone & Stanig, 2018).
Sudden changes in economic conditions, influence elite preferences and can challenge incumbent elites (Reuter & Gandhi, 2011). This is especially true for changes in trade, since many ascending elites rely on commerce to climb the social ladder. Hence, trade shocks have a large historical relevance (Scheve & Serlin, in preparation) and should not only matter to explain electoral behavior in contemporary democracies but also to understand elite fragmentation in electoral autocracies in the past.

Hypothesis 1.
A larger exposure to trade shocks increases the likelihood of a regional elite split.
The split of certain elites and the emergence of a new political party does not only depend on economic conditions. Political entrepreneurs consider not only the benefits-for example, more freedom to demand better suited policies or many offices-but also the costs of elite divisions-the efforts to mobilize the electorate (or to overcome institutional hurdles) to win seats (Cox, 1997;Tavits, 2008). Elites always consider the extent of uncertaintypolitical, economic, or institutional-before deciding whether to split elite unity and create a new political party (Lupu & Riedl, 2013).
The ability to mobilize voters is a widely acknowledged factor reducing the extent of uncertainty and it is connected to party formation (Kalyvas, 1996;Ziblatt, 2017); similarly, the existence of national agitators or agents of identity-based mobilization is associated with national revival movements across Europe during the 19th and 20th centuries (Hroch, 1985). Hence, it is important to focus on mobilization capabilities when trying to understand the logic of elite splits. Therefore, I hypothesize the following.
Hypothesis 2. Larger mobilization capacities in the hands of elites facilitate the emergence of regional party candidacies.
Beyond the main effects of trade shocks and political mobilization, I argue that both factors should reinforce each other. The main economic geography literature has always highlighted that economic and political factors are intertwined and it is important to take both elements into account (e.g., Beramendi, 2012). In my argument, what matters is not only the economic impact of the colonial trade shock or the political influence of mobilization agents. I argue that it is the combination of both elements what lies behind regional elite strategic decisions to break away from central elites. Regional elites in places largely exposed to the colonial shock required mobilization tools to be competitive in elections, while a lack of mobilization capacities made it less likely to consider engaging in elite fragmentation. The impact of an economic shock is contingent on the availability of political mobilization tools, and vice versa.
Hypothesis 3. Mobilization capacities magnify the effect of asymmetric economic differences across elites.
Overall, I consider that regional elites will have incentives to break their relations with central elites under certain specific conditions. The existence of asymmetric economic preferences is essential, but it will only be relevant when asymmetries are deeper and/or more salient, as it happens after a trade shock. This, coupled with the availability of mobilization tools in the hands of regional elites, should lead to the emergence of a center-periphery elite divide. Without relevant changes in the economic, political, or institutional conditions, and without the possibility to mobilize voters, regional elites would not engage in the formation of a new party. I will test my argument and specific hypotheses by analyzing the case of Catalonia in the early 20th century.

The Argument in Context
There are two key elements that make Catalonia a suitable case to study the origins of elite fragmentation. First, Catalonia was much more industrialized compared to other Spanish regions. Catalonia was, nonetheless, heterogeneously industrialized (Carreras, 2019); while some areas where highly industrial, others remained essentially rural. This allows me to leverage differences in economic preference divergence and exposure to economic trade shocks.
Second, efforts to build a shared and strong Spanish identity failed in the 19th century, partially due to an ineffective mass schooling system (L. Balcells, 2013). Although language or identity had not significantly divided regional and central elites until the 20th century, it was a visible and distinctive trait to be employed as a mobilization device if necessary. Whereas identity characteristics were quite homogeneous across Catalonia, the capacity of elites to activate voters on the basis of identity basis was not. There was variation in the presence of local leaders engaged in Catalan culture and language revival.
The previous elements assure enough internal variation within Catalonia to test whether economic grievances and mobilization capacities are related to the formalization of an elite split. Economic conditions and mobilization capacities should also explain the rise of regional elites in other regions-for example, Ireland, Bavaria, or Scotland-with identity-related and economic disparities between the region and the rest of the country. I focus, nonetheless, on Catalonia because the large internal heterogeneity within the region makes it the perfect case to test my hypotheses. In the upcoming sections, I briefly summarize the main economic and political characteristics of this territory.

Preconditions
Economic Context. The 19th century Spanish economic geography was heterogeneous: most areas remained almost entirely agrarian, while in a fewincluding Catalonia-industrialization was taking off (Prados de la Escosura, 2003). Catalonia consolidated its preeminent industrial position throughout the 19th century and by the end of the century it produced 40% of the country's industrial output and around 80% of the overall textile production (Nadal et al., 1988). 5 Textile industries were predominant in Catalonia and the cotton sector was the largest, followed by the wool sector (Nadal et al., 1988). Despite their shared industrial nature, both sectors had different characteristics. Almost the entire cotton production was based in Catalonia and depended on imported raw materials, while wool textiles were also manufactured in other areas of Spain since wool was produced in Castile. Moreover, cotton industrialists paid six times the amount of taxes paid by wool industrialists. Within Catalonia, each type of industrialization predominated in different areas and, as noted by Díez Medrano (1995), the specialization in different industry types influenced intra-elite relationships. Finally, different types of industries also meant different exposure to international trade.

Political Context
Since 1876, Spain was an electoral autocracy. Regular elections were held but these elections were neither free nor fair. Two notable parties, known as monarchist parties, systematically controlled over 80% of seats in parliament and perfectly alternated in power, thanks to fraud and electoral manipulation practices, which persisted even after universal male enfranchisement in 1890. 6 A majoritarian electoral system was in place for lower chamber elections and most MPs were elected in Single-Member Districts. Monarchist parties had a strong top-down control of the administrative electoral procedures (Villa García, 2016), which explains why non-monarchist candidates were almost never able to win seats. It was even difficult for non-monarchist candidates to win seats in other types of elections, such as subnational (provincial) assemblies, where constituencies were Multi-Member and candidates were elected through panachage . 7 Overall, this context represents a hard case to test the hypotheses of elite divisions, since winning seats against the will of incumbent elites in electoral autocracies is harder.
Political dynamics in Catalonia in the late 19th century did not differ from the rest of Spain, with large majorities of seats won by the monarchist party calling for elections. Bourgeois elites in Catalonia supported monarchist parties but remained relatively alienated from formal political representation. 8 Catalan elites conveyed their political influence mostly through economic societies or employers' organizations lobbies (de Riquer, 1977). However, the way Catalan elites engaged in politics changed by the turn of the century.

Elite Split: Key Events
In the late 19th century, there were two key events that had an important impact on Catalan elites and influenced the creation of the first regional political party in Catalonia in 1901. On the one hand, the increasingly structured organization of certain regional elites-mostly intellectual-along identity and language. On the other hand, the loss of the Spanish oversea colonies in 1898, which had important economic effects for Catalan industrialists (de Riquer, 1977). I will briefly describe the main historical events (see Figure 1) that influenced the decision of some Catalan elites' to politically split from existing political organizations and to promote a new regional elite party.
Identity-Based Mobilization. Catalan culture and language were being revitalized from the mid-19th century by a rising intellectual movement known as Renaixença [Rebirth], whose goal was to enhance the prestige of the Catalan language. It was a classical romanticist movement, their members were very keen on disseminating Catalan history and they defended the revival of Catalan self-government institutions. Although increasingly influential in society, intellectual elites were not strongly organized until the 1890s, when a new organization, the Unió Catalanista was created. This organization's goal was to unite all individuals committed to enhance Catalan culture beyond their other political allegiances, and to lobby political institutions in favor of Catalan identity. One of their first activities was to organize a series of assemblies to debate about Catalonia's future and its political organization, and to decide which issues had to be prioritized.
The first and most important assembly was celebrated in 1892 in the city of Manresa, and it gathered 237 delegates from all over Catalonia (Llorens i Vila, 1992). The assembly concluded with the approval of a document-praised by historians as a landmark for political catalanism-proposing the bases on how political relations between Catalonia and Spain should be structured. The Manresa meeting delegates were not secessionist, and most of them were liberal professionals and intellectuals, with very few industrialists among them. It was a sample of local elites who proselytized the revival of Catalan culture and political institutions, always within the Spanish monarchy. It is important to note that the meeting was not meant to lay the foundations for the formation of a political party but all the delegates were influential local leaders who were likely to promote any initiative that would enhance Catalan identity. Most delegates were not politicians, although some delegates would become prominent regional elite party leaders in the 20th century. Overall, we can think of the Manresa delegates as a network of activists promoting Catalan identity.
Historical Trade Shock. Later in the decade, in 1898, a critical juncture for intraelite relationships took place. The Spanish defeat in the Cuban War of Independence represented the loss of colonial possessions and "the complete collapse of the nation … [Not] only the institutions of the oligarchic parliamentary monarchy, but the very idea of 'Spain' came under criticism." (Álvarez Junco, 2002, 32). After the defeat, some elites-Catalan industrialists among themdemanded modernization reforms to "regenerate" the country but also to compensate the economic losses related to the sudden drop in colonial exports. The 1898 defeat had economic consequences and led towards a more general crisis that highlighted the need to introduce political reforms in Spain.
Colonial trade had been a key factor explaining the rise of Catalan industrialization. Preferential access to the colonial market was an effective compensation mechanism to facilitate the growth of Catalan textile industries given the low purchasing power of peninsular Spanish citizens. The access to the colonial market for Catalan industrialists ensured a larger market that was protected from international competitors through tariffs, therefore contributing to elite unity during the late 19th century (Nadal et al., 1988). As long as central elites supported protectionist policies facilitating commerce with the colonies and hindered the entry of textile products from other European countries, Catalan industrial elites had no incentives to split from central elites.
The loss of the colonies had large economic effects for certain economic sectors in Catalonia. Aggregate economic data reflect that the consumer price index (CPI) increased in Spain in the last years of the 19th century (Prados de la Escosura, 2003). Moreover, micro-level evidence from the city of Barcelona after 1898 has also proved a slight CPI increase after the war, which, combined with stagnated salaries, had important economic effects (Maluquer de Motes, 1999). The most compelling evidence of the economic effects of the colonial loss is the fact that taxpayers in industrial areas in Barcelona and surrounding areas engaged in a tax strike in 1899. This strike was a response to increases in taxation aimed to balance public accounts after the colonial loss and it clearly reflects the largely noted economic effects of the 1898 colonial defeat. Nevertheless, the colonial shock did not affect all economic sectors equally.
In 1895, the Catalan textile industrial sector exported around 25% of its production. However, while around 96% of all cotton textile exports were directed towards the colonies, the same figure for wool exports represented a mere 23.6% (Harrison, 1974;Dirección General de Aduanas, 1900). The 1898 colonial loss drastically reduced the demand of textile products from the colonies and forced textile producers-especially cotton ones-to try to relocate a large part of their sells in the Spanish peninsular market. Catalan wool industrialists were less affected than cotton industrialists, and Spanish agrarian elites-mostly landowners-even less so, given that agriculture products were not massively exported to the colonies. The 1898 defeat had relevant economic consequences and it was a catalyst for political realignments among those actors most affected by the colonial losses.
The Formation of a Regional Elite Party. To cope with the impact of the 1898 crisis, a "reformist" cabinet was formed with the initial support of the Catalan industrial bourgeoisie. However, when the new cabinet raised commercial and industrial taxes in order to balance public accounts, support among Catalan industrialists was largely lost and, as described previously, a tax strike was organized. This episode-as highlighted by this article's opening quoterepresented the first step taken by certain Catalan economic elites to break away from central elites and to do so through identity appeals.
Historians contend that, after being unable to influence Spanish politicseither through lobbying or through traditional political parties-industrial elites opted to organize a new political party with the support of the already organized intellectual intelligentsia (Ehrlich, 2004). The new party was called Lliga Regionalista-or Lliga-and it fits the definition of elite split presented previously: it was a new formal organization and some of their members had previously been affiliated with the two main Spanish-wide elite parties in the past. 9 The Lliga stood for elections for the first time in 1901 and it was a quite modern political party: it accepted individual membership and created a stable network of party delegations in multiple cities, the party was backed by a widely diffused newspaper, and it was able to organize large rallies (Molas, 1973). The Lliga expected to mobilize heterogeneous groups of voters-from upper and middle classes to rural peasantry-through appeals to Catalan identity and to impede fraudulent electoral practices by promoting large scale political mobilization.
Overall, the historical details presented above show how certain Catalan elites experienced an asymmetric economic shock after the 1898 colonial loss, which coincided with the appearance of a new network of identity entrepreneurs. All this resulted in the formation of an elite split through the creation of an elite-led regional party.

Data and Research Design
I explore the preconditions leading to a regional elite split by analyzing a dataset that combines electoral, economic, mobilization, and sociodemographic data. 10 "Replication materials can be found at Vall-Prat (2022)" The data is novel, it was compiled at the most micro-level possible (census-tract or municipality) and then aggregated at the constituency level.

Electoral Data
I employ election data at the constituency-level for the period between the creation of the Lliga in 1901 until a military coup put an end to the Restoration period and regular elections in 1923. The data include elections to the Spanish Lower Chamber and to Provincial (subnational) Assemblies, which had different characteristics-for example, district magnitudes and boundaries. 11 I rely on constituency-level electoral returns collected by historians, complemented with originally collected data. 12 The main dependent variable is a dichotomous indicator capturing whether any candidate of the Lliga was running for office in each electoral contest. 13 The dependent variable captures the elite split for each constituency. Figure 2 displays the number of regional elite candidacies per constituency between 1901 and 1923 and reflects that center and north-eastern Catalonia were the strongholds of regional elites. Finally, the analyses also include electoral variables capturing the type of elections, by-elections, and the magnitude of the electoral district.

Exposure to Colonial Shock
One of the main independent variables captures the economic impact of the 1898 crisis through an indicator of the colonial trade shock. The estimation of this measure combines two different sources and I employ procedures inspired by the Autor et al. (2013) shift-share instrument. On the one hand, I employ the change in exports-the variation in the values of exported products-to the Spanish overseas colonies at the national level between 1895 and 1900 from Dirección General de Aduanas (1900). 14 On the other hand, I capture industrialization tax records from archival data, to measure how relevant each economic sector was at the municipality level. 15,16 Tax records have been compiled for (almost) all municipalities in the earliest available year possible in each province, 17 and detail the taxpayers and the amount of taxes paid for non-agricultural economic activities. Once municipality-level data was available it was possible to aggregate taxation data at the constituency level.
The construction of the exposure to the colonial trade shock indicator is based on the following steps. 18 First, all typologies of exported products are matched to taxation classes. Since there is no perfect corresponding match between the two, most export product categories are linked to multiple taxation classes. 19 Second, I calculate the national level variation of exports towards the colonies for each product between 1895-the year the Cuba War started-and 1900-two years after the colonies gained independence. Third, I calculate exposure to the colonial shock (ECS) by each tax class (q) as a function of the different exported products associated to a specific tax class, as described in equation (1) ECS Δ p represents the variation in the values exported for each category product (p) between 1895 and 1900, Value p 1895 represents the value of all exports of product p in 1895, and Value q 1895 denotes the value of all products associated to each tax class q. These calculations give us a measure of exposure to the colonial shock per tax class. 20 Finally, a similar procedure is applied to calculate a weighed measure of exposure to the colonial shock for each constituency (j). The procedure is detailed in equation (2) ECS j ¼ X ECS q × Taxes q Taxes j ECS j represents the exposure to the colonial shock in each constituency and it is calculated by weighting the colonial shock exposure for each tax class (ECS q ) according to the relative size of taxes paid for the tax class (Taxes q ) over the total amount of taxes paid in the constituency (Taxes j ). 21 Figure 3 displays the geographic distribution of exposure to the colonial shock at the constituency level. 22

Identity Mobilization
The second pillar of the theory regards the mobilization capacity of regional elites. Along the lines of Hroch's (1985) contribution, the presence of local identity-committed leaders was a signal for regional elites regarding their chances of mobilizing voters through identity appeals. The 1892 assembly in which intellectual elites discussed about the future of Catalonia provides information on who these relevant local leaders were. The presence of these delegates was used as a shortcut to assess whether an elite split along identity lines would be sufficiently supported in each constituency. Large numbers of these agents of mobilization meant that previous efforts to increase the salience of identity in social debates had been made, and that the likelihood of electoral success was higher. Data on Catalan agents of mobilization in the 1890s from Llorens i Vila (1992) provide geo-referenced information regarding the municipality of origin-the locality where they lived or where they had ties-for all delegates attending the 1892 Manresa Assembly. Municipality-level data are then aggregated at the constituency-level. Identity-based mobilization capacity is thus proxied by the total number of delegates from each constituency that attended the Manresa Assembly. Since 20% of the delegates were based in the capital-city of Barcelona, I employ the logged number of delegates in the analyses. 23

Other Variables
Economic and Industry Indicators. In the mechanisms and robustness sections, I employ alternative measures to account for the exposure to the colonial shock. For instance, considering that municipality-level industrialization data are posterior to the colonial shock, I employ data on industries in Catalonia in 1861 (based on Jiménez Guited, 1862) to account for pre-colonial trade shock levels of industrialization. These data can be calculated separately for different types of textile industrialization (cotton and wool), that were differently exposed to the colonial shock to test the impact of the shock with data before Figure 3. Constituency exposure to the colonial shock. the shock. Similar data for 1931 are also available and allow me to interpolate industrialization characteristics around 1900.
The analyses also include controls for the overall economic activity at the constituency level through aggregate data on the global amount of industrialization taxes. Municipality-level tax data have been collected at (approximately) 5-year intervals and interpolated to obtain yearly data, which were aggregated at the constituency level. The logged value of overall taxes is employed as proxy for economic development.
Controls. The analyses include variables to consider other relevant social characteristics. These data include yearly interpolated census data accounting for total population, gender composition, literacy, marital status, and the percentage of people born in a different province-only available for the 1887 census.
Economic controls beyond industrialization taxes include an indicator for land ownership inequality, which accounts for the power of landed elites. 24 I also account for alternative political mobilization mechanisms through an indicator on the presence of landowner associations (Planas, 2006) and for the presence of peasant associations in the municipality (Pomés i Vives, 2000). Finally, the analyses include spatial controls such as the distance between the constituency's centroid to the city of Manresa. Table 1 summarizes the descriptive data.

Empirical Strategy
I analyze the origins of elite splits in Catalonia employing constituencyelection data between 1901 and 1923. I rely in a logistic estimation model in which the dependent variable is the presence or not of a regional elite candidate (Lliga) in each constituency (j) and election (t). The main independent variables capture the uneven exposure to the 1898 colonial shock (Colo-nialShock j , or CS j ) and the mobilization capacity of regional elites (Mobilization j , or M j ). The model also includes a vector of controls X 0 jt (election type, population, literacy rates…) and a time trend (λ t ) to account for time-variant characteristics that are constant across all districts. Since within-district changes are not of main interest and many of the explanatory variables are constant over time, the models do not include constituency fixed-effects. The model is summarized in equation (3) Lliga To provide more evidence that the colonial shock was the effective trigger of the regional elite split, I probe the plausibility of the shock by looking at different key aspects in the mechanisms section. For instance, I provide a rough test in which I compare the prevalence of different textile sectorscotton versus wool-that were differently exposed to colonial trade. I also look at other elements, such as relative timing with regards to the colonial shock or the type of elections. Although my design is not a difference-indifferences (DiD)-regional elites did not run for office before the shock-I examine whether the colonial shock is correlated with previous political dynamics. Along the lines of parallel trends assumption, differences across constituencies should be uncorrelated to their exposure to the shock to claim that it was the shock what made them politically different after 1898. Finally, I show the robustness of the findings by employing alternative variables and estimation models. Table 2 displays the results of logistic estimates assessing the likelihood to observe a regional elite candidate-the proxy for elite split-based on exposure to the colonial shock and mobilization capacities. The estimates account for other covariates and include time-trends and province fixed-effects to adjust for specific factors that might have affected constituencies over time or within provinces. According to the expectations, the constituencies that were more affected by the colonial shock were more likely to see a regional elite candidate Table 2. Baseline Results on Regional Elite Fragmentation. Logistic Estimation. DV: Binary, regional elite (Lliga) candidate; Controls: Election Type, Byelection %, Literacy, % not born in province, (log) Industrialization, Peasants Organizations, Landowners Organizations, distance to Manresa, and mean (log) population; (Std. Err.): Clustered at the District level. *p < 0.10, **p < 0.05, ***p < 0.01. Figure 4. Probability of Lliga candidate standing (95% CI). Based on model 3 in Table 2.

Main Results
running for office. Simple bivariate estimates relating colonial shock exposure and the presence of Lliga candidates are positive and statistically significant. 25 The effects reported in Table 2 are substantively large: in those constituencies least affected by the colonial shock, the probability of a Lliga candidate standing was around 30% and for those largely affected it was around 80% (Figure 4, left panel). Overall, the results confirm Hypothesis 1. The main effect for the mobilization variable in Table 2 is also consistent with Hypothesis 2: the effect of identity-based mobilization agents is positive and statistically significant. Differences are also substantive in size. The likelihood of the presence of a Lliga candidate almost doubles when the (log) number of attendants to the 1892 Manresa Assembly is at its mean compared to those constituencies with a minimal number of attendants (Figure 4, right panel). Regional elites were sensitive to the availability of mobilization agents when deciding whether to split or not from central elites.
Finally, the heterogeneous effect is positive and statistically significant. 26 This suggests that both elements-an exogenous economic shock and mobilization capacities-reinforced each other, in line with Hypothesis 3. As displayed in Figure 5, the impact of the colonial shock on the likelihood of regional elites standing for office was null for lower values of the mobilization variable. However, in those places where regional elites had more capacity to  Table 2. mobilize voters, the effects of the colonial shock on the likelihood of regional elites standing became positive.

Mechanisms
To prove that elite fragmentation occurs when regional elites have both the incentives (economic grievances) and the opportunities (mobilization capacity) to split from central elites, I present plausibility probes on the specific channels through which regional elite fragmentation should operate. I focus on three different aspects: constituency-specific industry composition, timing, and strategic political considerations based on different types of elections. Online Appendix C displays all estimates.

Industry composition
In order to confirm that economic grievances brought about by the colonial shock influenced regional elite decision to engage in elite fragmentation, I compare the relevance of two different textile industrial sectors (cotton and wool) on the presence of regional elite candidacies. Cotton was the textile industry most exposed to the losses of the colonial shock, while other textile sectors-like wool-were less affected by the decline in exports after 1898. According to Figure 6 (and Table C.1), the presence and relevance of cotton textile industries increased the likelihood of Lliga candidates.
While measures related to cotton textile industries in Figure 6 behave as expected in Hypothesis 1-3, placebo tests capturing the presence of wool and other types of textile industrialization are not associated with an elite split. This test shows that the constructed measure of exposure to the colonial shock is not reflecting the effect of industrialization per se, but the intensity of economic grievances affecting regional elites after 1898.
This evidence supports the idea that winners and losers from the economic shock responded differently to the 1898 crisis. In similar tests, constituencies where industrial exporting sectors were stronger and were more likely to see an elite split, while for non-exporting sectors the impact was null or even reversed (see Appendix E.6). Hence, incentives to engage in elite splits only materialized in those areas where industrialization was associated with economic grievances from the colonial trade shock.

Timing
Another way to show that exposure to the colonial shock was driving the elite split is by taking timing into consideration. The effects of trade shocks should not last forever, as political actors adapt their preferences and strategies over time. Hence, we should expect to see larger effects of trade shock exposure Figure 7. Heterogeneous effects by time periods.
shortly after the colonial loss. Figure 7 displays the results when splitting the sample by time periods (see also Appendix Table C.3). 27 Results show large and positive effects of colonial shock exposure on the presence of regional elites in the initial period. The effects vanish in the decade between 1908 and 1917 and reappear-albeit smaller in size-starting in 1918, probably because of a similar exogenous trade shock associated with a drop in cotton textile exports to World War I belligerent countries after the conflict. Table 2 reports findings for two kinds of elections that were subject to distinct dynamics. Lower chamber elections were first-order elections, and MPs in the Spanish Congress decided on important issues such as tax and tariff legislation. Provincial elections, on the other hand, elected subnational assemblies. These assemblies had certain spending capacities, but their political clout was clearly smaller. Although provincial assemblies were important for regional elites, provincial elections were less of a priority and their decisions to run in these constituencies should therefore be less influenced by exposure to the colonial shock.

Type of Elections
According to the theory previously outlined, we should expect that economic grievances would make regional elites more interested in being elected to the lower chamber. This was also, however, the most demanding type of elections because of the first-past-the-post system, and because there was more at stake for central elites. If regional elites were interested in influencing tax decisions and minimizing economic grievances, we should see larger effects of colonial shock exposure in lower chamber constituencies. Results in Table C.2 show that exposure to the colonial shock had an important effect on the presence of regional elite candidates in legislative constituencies but not in provincial ones.

Testing Pre-Treatment Differences
It is important for the credibility of the previous results to provide evidence that, along the lines of parallel trend assumptions, the treatment variables are uncorrelated with pre-treatment political outcomes. That is, that absent the colonial shock, regional elites would have stood for elections at the same rates across all constituencies.
The only way to show this is to provide evidence that key pre-treatment political indicators-that is, before 1898-were not affected by the economic consequences of the 1898 colonial shock. By focusing on pacific turn dynamics, 28 I construct an indicator that measures the percentage of elected MPs who were members of the ruling party calling for elections in all elections between 1876 and 1898-data from Varela Ortega (2001). Higher values of this indicator reflect higher alignment between Catalan elites and ruling elites before the colonial shock as Catalan elites conformed with Spanish-wide political dynamics.
The relationship between exposure to the colonial shock and pacific turn is weakly positive and not statistically significant at the 95% confidence level (Figure 8,left panel). If anything, the positive correlation reveals that constituencies hardly hit by the shock complied more with pacific turn dynamics, which should reduce the likelihood to see elite splits in these constituencies. Compliance with pacific turn dynamics represent a hard test for the theory, given that this indicator reflects the strength of elite alliances.
Appendix D shows evidence of the nonexistent relationship between exposure to the shock and two other pre-1898 political outcomes. On the one hand, there is no correlation between colonial exposure and the percentage of non-monarchist MPs elected between 1876 and 1923, which captures levels of political contestation. On the other hand, there is no relationship between the colonial shock impact and mean turnout between 1890 and 1898. 29 Finally, there is no statistically significant correlation between identity mobilization patterns in the 1892 Manresa meeting and exposure to the Figure 8. Colonial Shock versus pre-1898 Outcomes. colonial shock in 1898 (Figure 8,right panel). This is crucial to argue that both indicators are capturing different constituency-level characteristics. In other words, that exposure to the colonial shock is not capturing constituencies that were already more mobilized around identity lines. Moreover, identity mobilization levels are similarly unrelated to pacific turn dynamics before 1898 (Figure 11). Hence, the previous tests point out that colonial shock and identity mobilization indicators were uncorrelated with previous political outcomes.

Shock Validation Exercises
To disentangle whether my measure of exposure to the colonial shock is capturing the economic impact of the colonial trade shock, I present alternative ways to categorize industrial sectors. To this end, I classified the different industrial activities into exporting or non-exporting (more accurately, less-exporting) sectors based on the amounts exported towards the colonies. This measure should validate the colonial shock indicator and provide further evidence that it is a good proxy for the economic impact of the colonial losses. Figures E.2 to E.5 test the robustness of the interaction term when distinguishing between export oriented or non-export oriented sectors, and across specific industrial sectors. Results show the robustness of the measure by revealing that positive effects only hold for exporting industries-those losing the most from the 1898 shock-while for non-exporting sectors the effect is null or even negative. 30

Further Tests
Several robustness checks support the strength of the main findings presented in Table 2. 31 First, results hold when excluding elections in Barcelona, which had very idiosyncratic characteristics (Table E.2), and when excluding constituencies for which the construction of colonial shock exposure was based on fragmentary municipality-level data (Table E.3). 32 Second, I show the robustness of the variable accounting for regional mobilization agents. I employ alternative identity mobilization measurements and equivalent results are obtained when employing the absolute number of Manresa delegates, the percentage of localities with at least one delegate, or the number of delegates per ten thousand inhabitants in the constituency . Regardless of the indicator employed, the effect of the availability of mobilization agents is consistent with Hypotheses 2 and 3.
Third, I test whether the usage of different types of time controls changes the results. Results remain when replacing the time trend by year fixed effects (Table E.7). Moreover, robustness tests also confirm that the colonial shock exposure had the largest positive effects in the early 20th century and its impact decreased over time. Positive effects of exposure to the colonial loss were at maximum levels after the trade shock (see Figure E.6) but the positive effect was mitigated by the 1910s.
Fourth, results hold when considering not only support but regional elite electoral victories (Table E.8). This confirms the strategic evaluations of Lliga candidates, who decided to run mostly in those constituencies where they had larger chances of winning. 33 Fifth, I reestimate the models changing the unit of analysis and employing cross-sectional constituency data. I estimate OLS models with the dependent variable being the total number of elections in which regional elite candidates ran for office until 1923. In Table E.9, the number of observations drops substantially, but evidence supporting the expectations remains, at least for legislative constituencies, which is consistent with the mechanism described in the Type of Elections mechanisms section.
Finally, also using cross-sectional data, I show the robustness of the historical trade shock mechanism by employing different measures of industry specialization at the constituency level. Multiple alternative variables accounting for cotton industrialization are positively related to the number of regional elite candidates, which confirms the reliability of this indicator despite being constructed using industrialization data from the 1910s. 34 Overall, all robustness checks display a substantive effect along the lines of the effects shown in Table 2.

Conclusion
Intra-elite conflicts have important political consequences for democratization (Ansell & Samuels, 2014) or fiscal institutions (Mares & Queralt, 2020), but the origins and materialization of elite divisions have remained relatively unknown. The cleavage theory (Lipset & Rokkan, 1967), or theories of economic geography (Rickard, 2020) cannot explain the timing of elite ruptures because economic asymmetries and fixed social characteristics in most cases largely precede elite splits. Discontent elites need to possess both the incentives and the opportunities to engage in the uncertain enterprise of modifying the elite stability pact (Weingast, 1997). Incentives to split materialize when elite differences are intensified (Mor, 2022), and opportunities arise when disadvantaged elites have mobilization resources at hand (Kalyvas, 1996).
In this paper, I departed from the longstanding Rokkanian question (1970) regarding the origins of center-periphery elite tensions. The goal was to unearth the micro-foundations that lead certain regional elites to embrace the risky and uncertain process of elite division. I provided evidence that elite divisions along territorial lines occur when both economic grievances and the ability to mobilize the electorate are available to regional elites. By exploiting a historical trade shock, I have shown that exogenous economic shocks exacerbating economic differences can trigger elite divisions.
I have explored the formalization of elite divisions along a center-periphery dimension in early 20th century Catalonia, where the emergence of regional elites was not homogeneous across its territory (Molas, 1973). Neither Catalonia's economic geography, nor its distinct identity can account for the heterogeneous presence of regional elite candidates across constituencies. On the one hand, regional elites were unable to win seats in some highly industrialized areas; on the other hand, despite Catalan identity and language were homogeneous across Catalonia, elite fragmentation did not occur equally everywhere.
The elite split was associated with a large economic shock and with the availability of mobilization resources. Catalan regional elites mainly sought an autonomous political path in those constituencies that were largely hit by the effects of the colonial trade shock and where they could resort to identitybased mobilization. According to many historians (e.g., Harrison, 1974;de Riquer, 1977), the 1898 colonial trade shock intensified economic grievances and political and economic divergences between some regional and central elites. The colonial loss produced the necessary incentives for regional elites to break the existing equilibrium, but only in those places where mobilization capacities were already in place. The availability of identity-based agents of mobilization was necessary for industrial elites to successfully initiate a new political project that enhanced both the interests of intellectual and industrial elites.
The results show that elite divisions are rooted in economic heterogeneity, as advanced by Rokkan (1970). However, it is only when these disparities are suddenly experienced (or amplified) and elites can activate enough supporters that regional elites consider breaking away from central elites. These results underscore the need to better understand the origins of intra-elite conflicts and provide evidence on how (or when) certain political divisions are able to structure dynamics of political competition. Digging into the origins of different dimensions of elite competition can also contribute to fathom why certain institutional reforms were implemented in some countries and not in others as a consequence of intra-elite competition (Ansell & Samuels, 2014).
This work has focused on the center-periphery political divide, which is highly salient in contemporary research on decentralization and secessionist demands. Recent contributions have highlighted the role of economic factors to explain individual political preferences and behaviors along a centerperiphery political dimension. For instance, Hierro and Queralt (2021) have demonstrated that in contemporary settings voters employed in export-oriented economic sectors are more likely to support independence. Along similar lines, Rickard (2021) has shown how support for state-wide incumbent political parties declined when voters were exposed to offshoring events, while support for the only regional party in cabinet received larger support rates after these events. The conclusions in this paper are congruent with the findings on voters' political behavior. Regional elites in the early 20th century also responded to an economic trade shock based on their exposure to international trade dynamics: the more they depended on foreign markets, the more likely they were to emphasize their regional identity.
This paper contributes to the literature by improving extant understandings of the consolidation of political disputes along a territorial dimension, and detailing the circumstances under which regional elites reinforce their claims. The previous findings are consistent with new research on the origins of center-periphery divides and the rise of distinct political identities. For instance, Mor (2022) highlights the important role of unintended religious grievances for the development of a Catholic party in 19th century Prussia. Hence, the findings in this paper can be generalizable to other cases in which center-periphery tensions emerged-and cases where these divisions did not arise. The theoretical framework could be applied to many other cases where regional parties gained prominence, such as Ireland, Scotland, the Basque Country, or Flanders. For instance, while Ireland went through a series of famines during the 19th century, Scottish regional elites were not exposed to specific grievances and/or heterogeneous economic shocks compared to England. These factors can be relevant to understand why Ireland developed regional parties in the 19th century, much before the Scottish National Party was formed in the 1930s. Nevertheless, the importance of economic and mobilization factors can also be informative for contemporary settings. For instance, Siroky et al.'s (2020) contribution confirms the relevance of economic relations between the center and the region to explain the success of regional parties in Corsica.
Future work should explore whether elite divisions along other lines of political conflict were also influenced by shocks exacerbating elite-level tensions and the availability of mobilization tools. Additionally, new research on micro-level data for the early electoral success of regional elite parties could contribute to better understand the mechanisms at work in the consolidation of elite divisions and regional political parties. Further research is also required on the specific consequences of intra-elite divisions along center-periphery lines.
Regarding the origins of elite divisions, other factors could also account for the rise of center-periphery tensions. However, this work has shown that economic shocks and mobilization capacities are crucial aspects, and that elite tensions do not come out of the blue. A better understanding of the origins of intra-elite center-periphery tensions is the first step to further analyze the consequences of this divide in the regions where it materialized.
24. Based on the earliest data available, mostly around 1910s and 1920s and, only captured at one point in time-landownership structure was stable over time. 25. See Appendix Table E. 1. 26. Results are robust to Hainmueller et al. (2019) sensitive tests in Figure E.1. 27. Period categorization is grounded on work by historians (Ehrlich, 2004). Between 1901 and 1907, the Lliga was gradually developing and expanding, coinciding with the period in which cotton textile producers were adjusting to the colonial shock effects. After 1907, regional elites faced a crisis with internal party dissent, which was overcome after 1913, when a self-government political institution was created in Catalonia. Finally, after the end of the I World War, Catalan textile producers were hit (again) by a drop in exports to belligerent countries, economic grievances resurfaced, and industrialists faced a period of increased political and social instability. 28. As detailed in Political Context, this refers to the fact that most elected MPs belonged to the party calling for elections. 29. Between 1876 and 1890 elections had restricted franchise and turnout data is not available. 30. More details in Appendix E.6. 31. Empirical evidence in Appendix E. 32. See footnote 16. 33. Figure A.5 shows the geographic distribution of elected MPs. 34. Industrial activities did not disappear after the colonial shock and we should expect to see geographical continuities in industry composition (Nadal et al., 1988). This is confirmed in Figure 19, showing that cotton textile looms in 1831 were highly correlated with looms in 1931.