See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/242041956 Back in the Middle Ages? The paradox of identity: a microscope view of geographical and socio-economic heterogeneity in political representation in Article CITATIONS 0 READS 14 2 authors: Luca Solari University of Milan 33 PUBLICATIONS 167 CITATIONS SEE PROFILE Elena Perondi Politecnico di Milano 4 PUBLICATIONS 1 CITATION SEE PROFILE Some of the authors of this publication are also working on these related projects: Human resource management View project Evolutionary theories and change View project All content following this page was uploaded by Luca Solari on 25 December 2013. The user has requested enhancement of the downloaded file.
Back in the Middle Ages? The paradox of identity: a microscope view of geographical and socio-economic heterogeneity in political representation in Italy * Luca Solari University of Milan Via Conservatorio, 7-20122 Milan (Italy) tel. +39 02 50321183 fax + 39 02 50321165 email luca.solari@unimi.it Elena Perondi Catholic University Via Necchi, 5 20123 Milan (Italy) Tel. +39 02 72342427 Email elena.perondi@unicatt.it Draft, not to be quoted EGOS 21st Colloquium, June 30-July 2, 2005, Berlin, Germany. Sub Theme 10. Evolutionary Models of Organizations: New Perspective from International and Comparative Studies. * This research project was financed by FIRST funds at the University of Milan. The authors wish to thank Guido Legnante for his precious advice on the evolution of political representation in Italy.
INTRODUCTION As Aldous Huxley describes in his Brave New World, in a world well structured, diversity is attractive, but homogeneity eventually strikes back. The interplay of homogeneity and heterogeneity is a key issue in organization theory because it defines the nature of the very object we observe and analyze. At a first stage, views on heterogeneity and homogeneity define the boundaries between different approaches to studying organizations and originate paradigmatic boundaries. For example, supporters of differentiation relate the ability to create a sustainable competitive advantage to being different from competitors either because of a positional advantage or because of the resource base an organization possesses (Caves & Porter, 1977; Prahalad & Hamel, 1990; Wernerfelt, 1984; Porter, 1985). On the contrary, supporters of isomorphism maintain that institutional and societal pressures reduce change opportunities and limit allowed differentiation among organizations (Powell & Di Maggio, 1991; Scott, 1995). More insidious is the debate, which crosses the lines between theories and finds its realm within a specific theory as it has happened for Organization Ecology. If we adopt the view of environment as a constrained set of resources (Hannan & Freeman, 1989), the pattern of similarities (isomorphism) and differences (heterogeneity) among organizations encapsulated in that specific subset of resources, leads to the theory of density dependence (Hannan, 1989) based on the principle of competitive exclusion. If however we open up the door to a more complex interplay between forms, actors, and environment we need to focus our attention on the dynamic relation between isomorphism and heterogeneity, which could not be completely captured by density dependence. The theoretical formulation of Organization Ecology heavily draws on analogies to niche theory (Hannan & Freeman, 1983). The niche overlap that supports the density dependence argument requires heterogeneity within a given population to be low. But what might happen if we relax
this constraint? Could we still talk about an ecology of entries and exits from populations? Should we enlarge the theory spectrum to include explicitly heterogeneity within its theoretical and empirical boundaries? And in doing this should we consider that the generation of heterogeneity is not an exogenous process, but is dependent upon conscious agency by entrepreneurs, aiming at differentiating themselves from incumbents? In this paper, we demonstrate that population heterogeneity depends on socio-economic heterogeneity in audiences composition (Hannan & colleagues, 2004), but eventually takes a life of its own generating endogenous dynamics of density growth. We test our model on the population of political parties at regional level in Italy. Our results provide evidence of a complex relation between density, heterogeneity, and founding rates and support our theoretical formulation of heterogeneity dependence as a complementary process to density dependence. THEORETICAL BACKGROUND AND CONCEPTUAL FRAMEWORK The definition of the basic unit of analysis of Organization Ecology, i.e. population, clearly provides evidence of the interplay between heterogeneity and isomorphism. A population of organizations is a group of organizations that share some relevant organizational properties, i.e. goals, forms of authority, core technologies and marketing strategies. (Hannan & Freeman, 1989). Drawing on niche theory in biology, homogeneous sets of organizations (populations) are characterized by a common dependence on the material and social environment (niche). This common dependence from the external environment favors isomorphism and establishment of boundaries (Hannan, 1979). Nevertheless, across time these boundaries can blur and originate new populations, which can either be created as a consequence of processes of competition among sub-populations (Carroll, Hannan, Peli & Polos, 1997) or legitimate action by new entrepreneurs benefiting from exogenous innovation (Aldrich, 1999)
The sum of these processes of socio-economic innovation originates heterogeneity in organizations, which are rapidly structured into new forms and coded as legitimate actors, enforcing cognitive barriers into audiences. Therefore, heterogeneity among populations of organizations at the interpopulation and community level (Hannan & Freeman, 1977, 1989; Carroll, 1985; 1997; Baum & Singh, 1994) can be considered a consequence of the existence and enforcement of cognitive and social boundaries, which allow organizations to benefit from external and internal taken-for-grantedness and legitimation. The processes that operate in determining the boundaries among different populations have received scant attention (Barnett, Woywode, 2004), but are central to the development of Organization Ecology. The definition of boundaries among forms is central to theorizing in Organization Ecology, though by no means an easy task because different social and economic processes may take place at different levels. Considering geographical scope, for example, Carroll and Hannan (1995) found that legitimation processes cross political boundaries and operate at a broader scale than competition processes. Likewise, Lomi (2000) claims that standard density-dependent model is strongly supported at more aggregate levels of analysis, but fails to explain more finegrained mechanisms. A more detailed analysis of patterns of geographical heterogeneity are found in Baum and Mezias (1992). Baum and Mezias (1992) measure localized competition in the Manhattan hotel industry by the Euclidean distance between a focal organization and all the other organizations in the population. Heterogeneity as measured by this distance is found to affect failure rates of Manhattan hotels. When a focal hotel is similar to its competitors, it experiences higher competition and lower survival chances. The study, though, considers only competitive effects and does not take consider the impact on the legitimation process. A part from geographical heterogeneity, the most notable contribution in Organization Ecology is certainly the analysis of the effects of patterns of intra-industry heterogeneity in Carrol's resource partitioning model (1985). The resource partitioning model introduces concentration
among generalists as a key factor affecting life chances of specialist and generalist organizations. The model operates in a market where consumers are heterogeneous, price competition is weak and production or marketing activities exhibit economies of scale (Carroll & Hannan, 1995). Under these conditions, large generalists organizations compete to occupy the center of the resource space (i.e. the market), allowing peripheral resources to be available for specialist forms. Carroll introduces the term "partitioning" because, in concentrated markets, specialists and generalists rely on a different resource base. Strong empirical evidence of the resource partitioning model (Carroll, 1985; Barnett & Carroll, 1985; Freeman and Lomi, 1994; Carroll & Swaminathan, 1992) has been provided and many extensions are being analyzed (i.e. status based competition Podolny, 1998). As boundaries eventually need to blur to originate the world we can see, the role of heterogeneity in Organization Ecology has long been a central concern, because of theoretical (Lomi, 1995) and empirical issues (Peterson, Koput, 1991). We can distinguish two aspects of heterogeneity. First, heterogeneity can result from unobserved characteristics of organizations, which interplay with population dynamics. In such a case, we are confronting with the risk of misspecification of forms. At this level, one common way out has been to define categories (i.e. subpopulations) and analyze them as interdependent observable, attributing observed organizations to discrete groups. Second, heterogeneity can result from unobserved variation in the process of interplay between environment and forms when forms are specified adequately. At the processes level, several efforts have lead to consider spatial heterogeneity and location dependence (Carroll & Wade, 1991; Hannan & Carroll, 1992). In his research Lomi (1995; 2000) reconciles the two aspects of heterogeneity and shows that organizational populations are internally differentiated and that vital rates vary systematically across heterogeneous segments of the population (Lomi, 1995). We argue that within ecological processes, heterogeneity in forms (or organizational diversity) is not only a consequence of selection pressures, but the material on which those processes operate.
Organizations though following sometimes a blueprint are created diverse. The resources they use, the processes and roles, their mission, their very identities need to be diverse to appeal to audiences from which their survival depends. If they are to form the blueprint for a new organized activity to be established (i.e. a population), they need to provide a different value from incumbent populations and if they are to follow a blueprint within a population, they need to make a competing claim to incumbent organizations. Therefore, heterogeneity observed in existing populations of organizations (and in organizations within populations, as well) is the outcome of a much greater heterogeneity in terms of solutions to given social, economic, and political needs that has been exposed to selection processes (see Ruef, 2000). The processes by which heterogeneity in environment operates on organizations is a central constituency of Organization Ecology if we follow recent theorizing (Polos, Hannan, Carroll, 2002; Hannan, Polos, Carroll, 2004). Forms can be interpreted ad a set of codes enforced by specific internal and external audiences. Violating a code exposes to the risk of punishment in the terms of a reduced legitimation and therefore a loss of viability of a single organization. However this recount does not exclude the instance in which violating a code exposes to a prize because the organization is able to gain resources and legitimation from a previously excluded or neglected audience. Audiences are not fixed, they are subject to social processes and vary both in composition and in time. Any theory of organizational populations, therefore, needs to be complemented by a theory that analyzes how the interplay between forms and audiences unfolds and produces innovation in the composition of the organized world. Clearly speaking such a theory needs to take into account two aspects. First, it needs to establish a clear link between heterogeneity in audiences constitution and heterogeneity in organizational forms. Second, it needs to demonstrate that population dynamics are linked to heterogeneity as much as to density (or to be more extreme exclusively to heterogeneity with density dependence as a special case).
These processes are difficult to capture empirically and this might explain why most research ended up with a bias toward density dependence. However we think that if we were able to find an organizational population, which possesses the following characteristics: - an identity defined at the organizational level (or group) as well as at the population level; - a landscape of audiences potentially finely differentiated; - a strong interdependence between audiences needs and forms of organized activity, we could demonstrate how dynamics at the population level might hide a much more heterogeneous picture than commonly has been accounted for by existing research. In the following section we describe a population that enjoys these characteristics and that has been used for testing our hypotheses. REGIONAL POLITICAL PARTIES OVERVIEW The national political scenario Italy is a young democracy, established right after the end of WWII with a close-count national referendum, which banned monarchy. The specific nature of this country, characterized by a strong opposition communist party and a very heterogeneous geography, originated a highly specific political context. A sign of this is instability (between 1948 and 1973, Italy had twentysix governments) which was tempered by the fact that the Christian Democratic (DC) party, dominated during much of the postwar period and provided continuity and relative stability. Center-left coalitions dominated by the DC continued to hold power until 1983, when the republic's first Socialist-led coalition took power. The continuing sluggishness of the economy caused the institution of an austerity budget, which included tax increases, service cuts, and wage adjustments. The Socialist coalition led the government for four years, until the elections in 1987
that restored the primacy of the Christian Democratic party. During the next four years three different premiers, all belonging to the DC, followed one upon the other. In 1991 the traditional political scenario started changing and the Italian Communist Party changed its name into the Democratic Party of the Left with the spin off of a new political subject, Communist Refounding aiming at the traditional communist ideals and in clear opposition to the perceived right-turn. Despite the beginning of decay in the historical equilibrium among political powers, in the 1992 elections the Christian Democrats barely maintained their coalition with the Socialists, the Liberals, and the Social Democrats. Corruption probes begun in 1992 and led to the arrest of hundreds of business and political figures and the investigation of many others, including several party leaders and former premiers ( Tangentopoli ). From 1992 to 1997, Italy faced significant challenges as voters disenchanted with past political paralysis, massive government debt, extensive corruption, and organized crime's considerable influence demanded political, economic, and ethical reforms. In 1993 referendums, voters approved substantial changes, including moving from a proportional to a largely majoritarian electoral system and the abolishment of some ministries. The shift from a proportional to majoritarian voting system, with the requirement to obtain a minimum of 4% of the national vote to obtain representation, also altered the political landscape. Party changes were sweeping. The Christian Democratic party dissolved; the Italian People's Party and the Christian Democratic Centre emerged. Other major parties, such as the Socialists, saw support plummet. In this chaotic situation, when it seemed highly likeable that the former communist party could easily win the elections, a new populist, and free-market orientated movement, Forza Italia, gained wide support among moderate voters, favored by the high visibility of the tycoon that promoted it, Mr Silvio Berlusconi. Meanwhile the existing neo-fascist party, Italian Social Movement took an unprecedented turn and became the National Alliance, which was embraced
by several former right-wing Christian Democrats. A trend toward two large coalitions (one on the center-left and the other on the center-right) emerged from the April 1995 regional elections. For the 1996 national elections, the center-left parties created the Olive Tree coalition while the center right united again under the Freedom Pole. The center-right coalition won with the important support by a controversial, autonomist party (Lega Lombarda), to discover soon after that the balance could not be maintained. The following elections brought to power the centerleft coalition, which in turn broke out in 2000 leading to a new vote. The May 2001 elections ushered into power a refashioned center-right coalition dominated by Berlusconi's party, Forza Italia. The Olive Tree coalition now sits in the opposition. This emerging bipolarity represents a major break from the fragmented, multi-party political landscape of the postwar era, although it appears to have reached a plateau, since efforts via referendums to further curtail the influence of small parties were defeated in 1999 and 2000. The evolution of regional electoral system State, Regions, Provinces, and Municipalities constitute the Italian Republic. The autonomy of these jurisdictions is an important feature of the Italian political and institutional system. The Constitution of 1948 asserts that the Regions, as well as the Provinces and the Municipalities, are autonomous entities with their own Statutes, powers and functions (art. 114). Five regions, Friuli-Venezia Giulia, Sardinia, Sicily, Trentino-Alto Adige and Valle d Aosta enjoy particular forms of autonomy, according to their special Statutes adopted by constitutional law. The Constitution also identifies the branches of government of the Regions, which are the Council, the Cabinet and its President. While the Council exercises the legislative power and all other functions granted to the Region by the Constitution and by law, the Cabinet is the executive branch of regional government.
Yet, the rules that regulate the functions and the mechanism to elect or appoint such bodies of government have changed over our sample period. Until the reform passed in February 1995, voters elected the Council under a proportional system. The number of legislators varied depending on the regional population 1 and they held the office for a 5-year term. The Council appointed the Cabinet, which was composed by the President and a certain number of members, usually (but not everywhere) called assessori. The changes occurred in the political and institutional scenario in the early 90s and the difficulties in reaching stable governments led to an important reform of regional and local governments electoral system. The reform law has modified both the electoral system and the tenure length of regional legislators. Starting from 1995 the mechanism to elect the members of the regional Council switched from a pure proportional system to a mixed one. Specifically, 80 percent of the legislators are elected on the basis of provincial competing lists (art. 1, sec. 2) and the remaining 20 percent by a majority system on the basis of regional competing lists (art. 1, sec. 3). A premium for the majority in two steps is also introduced, such that the absolute majority of the legislators will be held by the coalition linked to the regional list that has obtained the relative majority of the votes 2. Furthermore, the law reduces the tenure length of the Council from five to two years when during the first two years the relationship of confidence between the Council and the Cabinet breaks down. 1 On the basis of the law passed in 1968 (n. 108) the members of the regional Council are 80 for the regions with more than six millions of inhabitants; 60 for the regions with more than for millions of inhabitants; 50 for the regions with more than three millions of inhabitants; 40 for the regions with more than one million of inhabitants and, finally, they are 30 in all other regions. This relationship between regional population and legislature size makes irrelevant possible endogeneity between regional government spending and legislature size (according to the argument that says that larger government spending needs more legislators because of increased specialization. 2 To allow for this premium, the number of legislators can be increased if votes do not give such a majority when translated into seats.
Another important change happened in 1999 when Parliament approved a constitutional law which modified the art. 122 (sec. 1). It states that the President of the regional Cabinet is elected by universal and direct suffrage, unless when the regional Statute establishes otherwise, and that he appoints and dismisses the members of the regional Cabinet. The constitutional law of 1999 also gives to the regions the opportunity to write their own statutes (art. 123). The statute determines the form of government and the fundamental principles of the organization and the functioning of the Region, in accordance with the Constitution. In other words, the Regions can choose their own form of government and electoral rules, within some boundaries. In particular, they can set the number of legislators. Twelve out of twenty regions plan to increase the number of legislators. This change will be effective from regional elections in 2005. Parties competition and coalition before the transition The main characteristics, that remarkably influenced the action and stability of regional government until 1992, were the structure of electoral competition and the coalition policy pursued by major parties in the regional contexts (Vassallo & Baldini, 2000). Regarding our analysis, during the twelve-year long period from 1980 to 1992, we can pick out a restricted group of national parties playing (which were playing) the leading role in the local scenario. On one hand there was the historical anti-communist coalition of four traditional parties (the Christian Democracy, the Social-Democratic Party, the Republican Party, and the Liberal Party), and on the other we can find the Communist Party (after Democratic Party of the Left). In the middle of this sort of polarization, a central role was played by the Socialist Party, which held the balance between a centre-left alliance or a left one. Based on the specific strength relations and political competition structures set up between these leading parties, Vassallo and Baldini (2002) single out four different groups of Regions:
a group characterized by a communist dominance (Emilia-Romagna, Toscana and Umbria); a group where it was possible, or advisable, to leave the Christian Democracy out of the majority because of the distance between the anti-communist coalition (the Centre ) and the Communist Party was moderate (Liguria, Lazio, Marche, Piemonte); a group where the Christian Democracy polled a lot of votes, consequently it was very difficult to form a government without its support (Lombardia, Puglia, Calabria, Campagna); a group in control of a significant Christian Democratic majority (Abruzzo, Basilicata, Veneto). In summary, there were two distinguishable clusters of Regions: the first consisting of Regions with a clear political connotation (the red Regions and the white ones), and the second made up of more disputable Regions. The former could have been characterized by a competitive dynamic, but at the end prevailed a consociational approach and the two different coalitions decided to cooperate in a larger majority of government lowering the political competition within Regions. Table 3 represent the regional mean in the number and births of parties and the minimum and maximum number of births among Italian Regions. The data reflect the stability of the regional parties population: the births are assembled closed to electoral events (1980, 1985 and 1990) and the new parties are mostly local and small sized, with two relevant exception: Lega Lombarda (1984) and The Pensioners Party (1983 in the most of the Regions). [Insert Table 1 about here]
The electoral reform and the majority system The Tangentopoli effect and the crisis of the traditional political system had direct consequences also on stability and governability of regional executives. During the two-year period 1993-1994 nearly all the Regions changed their majorities and councils. However, radical changes started in 1995 when the first election with the new electoral system took place, just a couple of week after its introduction. A lot of Regions alternated their governments and new dynamics of political competition were designed. In two years, starting with the break-up of Christian Democracy at the beginning of 1994 and closing with the split of Christian Union of the Center from one of the two parties emerged from Christian Democracy decay (Christian Democratic Centre) after the election in 1995, the whole political scene, both at national and regional level, turned into something completely new. The traditional parties disappeared and almost all the parties running for the electoral competition were new (or had new names): in 1995 we have an average number of births of 8,07 out of an average number of parties of 7,26 and a minimum number of births of 5 per Region (table 2). Notwithstanding the parties dynamics, after the institutional transition, were still centered on electoral events, we can observe an outstanding number of births also in the five years between the two elections (1995-2000) and after 2000. These dynamics are in part due to national adjustment and reconfiguration. [Insert Table 2 about here]
Further data are relevant for the correct interpretation of the political transformation: the number of parties and the effective number of parties 3 (ENP), an index that represent the political system fragmentation. Comparing the data in table 3 with the mean value of 4,4 (ENP-votes) drawn in 1990, after the political impact of the success of Leghe parties, it emerges an increase in the ENP value. In fact, after the introduction of majority system and direct election of the President, there was a clear bipolarization of political competition strictly related to a relevant fragmentation within the two coalitions (D Alimonte & Bartolini, 2002). [Insert Table 3 about here] HYPOTHESES We are driven by a lot of reasons to search for a new specification of heterogeneity in Organizational Ecology. First, the widespread adoption of a discrete distribution of organizations among forms is clearly a second-best choice. When heterogeneity within a population is high, this allows for (and to some extent requires) clustering. Clustering can be obtained through identification of specialist and generalist organizations or according to differences in technologies (Barnett & Carroll, 1987), legal form and ownership structure (Freeman, 1990; Rao & Neilsen, 1992) and size (Haveman, 1993). Clustering, though, shifts the implicit homogeneity assumption from the population level to the sub-population level but does not solve the problem. Second, considering niche width theory (Hannan & Freeman, 1983; Hannan & Freeman, 1989) specialism and generalism are not discrete, but continuous attributes. The niche width of a (sub)population is the variance of its resource utilization (Hannan & Freeman, 1989), clearly a 3 The ENP value can be calculated regarding the percentage of votes or the percentage of seats, the two formula are: = 2 1 % votes ENP votes and ENP seats = 2 1 % seats
continuous attribute. Therefore, dichotomization in specialist and generalist forms is just an empirical artifice. Third, the discrete setting of boundaries within a population is in contrast with a long-term evolutionary view. The discrete distribution in specialist and generalist forms over long time periods does not consider that population heterogeneity might change over time. If heterogeneity is considered variable over time, clustering could not be appropriate at all stages of the life of a population. Theoretically, during periods of transition and revolution, the boundaries defining forms are likely to become clouded and ambiguous (Polos, Hannan & Carroll, 1999) rendering a discrete distribution arbitrary. Even if social actors use a limited range of types to define forms that does not mean that the world of organizations is really discontinuous (Hannan & Freeman, 1989). Therefore, clustering organizations according to both few variables or a wider array of attributes is anyway logically wrong because heterogeneity is a continuous variable. Fourth, in the analogy with evolutionary biology, organization ecology is interested in ascertaining which factors affect vital rates of organizations within a population. The degree of similarity (or heterogeneity) among individuals within a population is clearly a measure of variety, which is one of the four forces of any evolutionary theory (Aldrich, 1979). While in the natural world variety is randomly generated by mutation, in the business world it is a consequence of deliberate action. For these reasons, heterogeneity must be considered explicitly as an attribute of a population (Hannan & Freeman, 1989). Different dynamics of competition and legitimacy may result in different levels of heterogeneity within a population or across populations. Because of the implicit assumption of niche overlap and homogeneity laying at the roots of the density dependence model, different conditions of heterogeneity might influence its validity, and call for a theoretical extension of organization ecology. In our perspective, heterogeneity has to be measured at the population level as a continuous variable to assess whether it affects entry rate of members of a population in any stable way. We
assume that density-dependence model (both in his legitimation and competitive effect) is lacking in explaining founding dynamics because it doesn t take into consideration measures of heterogeneity. First of all, we argue that population heterogeneity stems from audience heterogeneity, assuming as audience sets of agents with an interest in a domain of activity and some control over resources (material and symbolic) that matter for those attempting to carve out positions in the domain [ ] (Hannan, Polos & Carroll, 2004). In respect with Hannan and colleagues definition we reject the assumption of relatively homogeneity in audience s tastes, in fact we suggest that differences in some basic characteristic of the audience (i.e. audience heterogeneity) lead to diversity in agents needs and tastes. A widening in the spectrum of organizations features, belonging to a given population, reflects an enlargement in audience heterogeneity. HP 1. Population heterogeneity is positively related with audience heterogeneity. Furthermore, a new political party in order to run for regional (and also political) electoral competitions is supposed to gain the support of a share of its electoral body. In fact the admission of a new party to elections in a given constituency is made conditional on the collection of a number of constituents signatures proportional to the total number of voters in that constituency. Virtually, a party needs audience legitimation before it can be considered a new political entity. Consequently we expect an increase in founding rate whenever a new necessity arise and create an unsatisfied request from audience (demand pull effect). HP 2. Audience heterogeneity will positively affect founding of new parties. Starting from assumptions based on the first hypothesis, we claim that the definition of legitimation as taken-for-granted, which is one of the starting point of the density dependence model (Hannan, 1989; 1991), is based on a static vision. We propose to look at the concept as a dynamic process of individual and collective evaluation. Legitimation has no specific bureaucratic-legal implications, it is a broader social phenomenon that takes time to develop: legitimation of a new form is a
combined function of its persisting presence on the bureaucratic landscape, which helps to disseminate the image of the form (constitutive legitimation), and the extent to which organization gains acceptance by the powerful social actors (sociopolitical legitimation) (Dobrev, 2001). Moreover the degree of resource dependence will moderate responsiveness to isomorphic pressures and determine the extent to which organizations seek to differentiate themselves (Ogata, 2004). Generally speaking, the problem lies in coupling heterogeneity with different kinds of needs and, consequently, it s necessary to take into account the working of the social system (Thompson, 1967). Legitimation effects stem from the common dependence of organizations on societal and cultural goods, like recognition, acceptance, accountability (Solari and Rossi, 2000). Based on previous assumptions, our hypothesis is that a new birth is the response to a new or unsatisfied need expressed by a new or changed social group. In order to follow and to satisfy audience s needs the new organization will be different from the existing ones and, therefore, it will increase population heterogeneity. Considering the symmetry of the relation, we can assert that an increase in the population heterogeneity brings to an increase in the birth rate. This assumption, seemingly in opposition to intuition, can be explained thanks to a disputable effect : generally speaking, a high heterogeneity is indicative of the level of fragmentation and differentiation in audience s needs, so there are more opportunities to identify and fulfil a niche of necessities without representation. On the other hand we suggest that a low degree of heterogeneity within a population is due to high polarization and homogeneity in audience s tastes. In the former situation the population needs are already well represented by existing organizations and the coming ones have restricted degrees of action. HP 3. Population heterogeneity will positively affect founding of new parties
METHODS Data We conduct an analysis of Italian regional parties ecology between 1980 and 2003, based on two classes of data: socio-economical data and election data. The socio-economical data are used to evaluate the population heterogeneity, including both economical and social aspects. The most important source for this kind of data in Italy is ISTAT (Italian Institute of Statistics), which produces annual reports on socio-economical variables. We derive the main economical data from Local Economical Accounts: per capita GDP, income per dependent worker, and per capita expenses. As social references, we singled out variables that represent significant indicators of religious level, information diffusion and working conditions of Italian regions: percentage of religious marriages on total marriages, number of TV licenses, percentage of working population on total resident population, percentage of workers employed in Services sector on total working population, unemployment rate, number of working conflicts and of strike hours per worker, and the percentage of workers involved in conflicts on total working population. The aim to include also statistics regarding educational level (e.g. percentage of high-school and/or university graduated, and illiteracy level) has been inhibited by the unavailability of this kind of data per year at regional level. These data can be extracted only from ten-yearly National Census, but this would lead to a mix between continuous and discontinuous time series. The data regarding social aspects stem from Statistics Yearbooks published by ISTAT. The regional parties election data are extracted from Istituto Cattaneo, an institute that collects data and perform analysis on Italian elections since 1946. We decided not to use Ministero degli Interni (Home Affairs) as a source, because it provides the electronic format of election data just since 1995 (for regional elections) and 1994 (for national elections). Using only election data, we should assume that a new party foundation matches with the year of the first election
where that party is present. This assumption leads to a bias of the dynamics of National parties that runs for Regional elections, concentrating all the births in election years: the elections would become the main factor influencing entry rate. On the contrary we could assume that national dynamics affect local ones. For example, the Italian Communist Party (PCI) changes his name into the Democratic Party of the Left (PDS) with the spin off of Communist Refounding (RC) in 1991. Because we have regional elections in 1990 and in 1995, not considering national dynamics at local level, we should fix the end of PCI and the birth of PDS and RC in 1995. On the other hand we shall consider, also at local level, that in 1991 the regional and municipal federations of the PCI became representatives of the new PDS or RC. Based on those considerations we decided to collect data on the history of national party population using political literature (Galli, 2001; Dalla Porta, 1996; Corbetta & Parisi, 1988) and official websites. Measures for Analysis Independent variables Our aim is to focus on the formation of regional parties in relation to their audience and to some attributes of their population. In order to measure the independent variable we collect data on regional parties births per Region and per year. Parties heterogeneity One of the independent variable for the organizational founding rates model is the parties heterogeneity in a given Region. It is calculated using the Simpson diversity index, that defines the diversity as the probability that two randomly selected individuals in a community belong to different species (Maignan, Ottaviano, Pinelli and Rullani, 2003). Given a community: C = {s; π 1, π 2,, π S }
where s is the total number of species and S π i = N i / k = 1 N k is the proportion of all individuals that are of specie i, i.e. the probability that an individual belongs to specie I, the Simpson diversity index (Δ Simpson (C)) is: S 2 Δ Simpson (C) = 1 - k = 1 π k The Simpson diversity index is a measure that accounts for both species richness and weight S inside the community: k = 1 π 2 k is influenced either by the number of species or the equitability S of percent of each specie present (for a given specie richness, k = 1 π 2 k, will decrease as the percent of each specie becomes more equitable). An alternative measure of diversity is the Shannon-Weaver index, that gives more emphasis to the rare species. We decided to use the Simpson index in order to give much less weight to political coalitions with few parties (rare species in our study), because, generally speaking, they are formed by marginal local parties, most of them living for one or little more elections. In our case, the species are defined into a matrix with two dimensions: coalitions and national/local representativeness. Starting from 1995, after the introduction of the electoral reform (majority system), the electoral system is naturally based on the parties aggregation into coalitions, so the species are simply defined as the coalitions that runs for elections. Before that date, we classified as part of Center coalition the five traditional government parties ( Pentapartito ), while as part of Right and Left coalitions the historically sided parties. Other coalitions have been defined considering the represented local or specific interests. The division between national and local parties level is based on the following criteria: if a party runs for a national election (in the period of our analysis) or if it runs for more than four regions in regional elections, it can be considered as a national party.
Socio-economical heterogeneity The independent variables in the analysis of founding rates include regional heterogeneity of the selected socio-economical variables: per capita GDP, income per dependent worker, per capita expenses, percentage of religious marriages on total marriages, number of TV licenses, percentage of working population on total resident population, percentage of workers employed in Services sector on total working population, unemployment rate, number of working conflicts and of strike hours per worker, and the percentage of workers involved in conflicts on total working population. The heterogeneity of the socio-economical variables under analysis is calculated per year and per region. The standard among organizational demographers for measuring group heterogeneity is the coefficient of variation (defined as the standard deviation of a variable divided by its mean) (S rencsen, 2002), even if this measure is sometime misused. In our case, we excluded the use of the coefficient of variation mainly because it would give a unique value per year, while we need a measure per region per year. We choose, as a measure of heterogeneity for socio-economical variables, the following ratio: Δ it = X it X σ t t Where: X it is the value of considered variable for region i in year t; X t is the weighted average of regional values in year t, using regional resident populations as relative weights; σ t is the standard deviation for year t. The numerator appraises the diversity between different regions included in the sample. The ratio with standard deviation is functional to weight the diversity with a measure of scale and of sample distribution, obtaining a relative diversity measure expressed in percentage.
Empirical Specifications The sample of analysis comprehends 14 of the 20 Italian Regions: we excluded the 5 Regions under special statute (Friuli Venezia Giulia, Sardegna, Sicilia, Trentino-Alto Adige and Valle d Aosta), because of their high local autonomy and their particular political and electoral systems, and Molise because its election events suffer form time lag with respect to other Regions. In order to manage the great amount of data and to make relations between variables clearer, we decided to perform a factor analysis. The first analysis reveals that two (number of TV licenses and percentage of workers involved in conflicts on total working population) of the twelve socioeconomical heterogeneity variables, are characterized by a high value of uniqueness, consequently we excluded them. The second test, a factor analysis with varimax rotation, allows us to group variables into three significant factors (Tab. 4): [Insert Table 4 about here] factor 1, the most representative, groups 6 variables (heterogeneity of: per capita GDP, income per dependent worker, per capita expenses, percentage of working population on total resident population, unemployment rate, and percentage of religious marriages on total marriages), and it represents diversity among Regions with respect to Economic Development; factor 2 comprehends the heterogeneity of percentage of workers employed in Services sector on total working population and it is indicative of regional differences in the Expansion of Service Industry;
factor 3 encapsulates two variables (heterogeneity of number of strike hours per worker, and of percentage of workers involved in conflicts on total working population) and it represents diversity in Social Conflict. Figures 1a and 1b map the distribution of factors 1 and 3, while figures 2a and 2b show factors 1 and 2, respectively in 1980 and 2003. The first evidence is that there are no clear patterns of dependency between factors, so independent variables seems not correlated between each other. Secondly it is pretty evident a polarization process from 1980 to 2003: for each of the three factors we can see an increase in the dispersion of 2003 values, with respect to 1980 ones. [Insert Figure 1a, 1b, 2a and 2b about here] Given our convictions that population heterogeneity has significant relations with audience needs and tastes and with changes and evolutions of those, we test out hypothesis 1 using a linear regression that includes the three factors representing audience heterogeneity, the regional density and some time variables: HET ( t) htime 2 = a + bedt 1 + cesi t 1 + dsct 1 + en t 1 + ftan + gtime + (1) Where TAN is a dummy variable coding 1 for years after 1993 (i.e. after Tangentopoli and the consequent electoral reforms) and 0 before. Table 5 reports interesting results. First of all, political parties population heterogeneity shows a high, positive and significant relation with two of the three audience heterogeneity factors: economic development heterogeneity(ed) and expansion of service industry heterogeneity (ESI), while social conflict heterogeneity (SC) seems not to have any relevant impact. Second, regional density (N) at t-1 positively affects parties heterogeneity and, finally, we can observe a positive first order and negative second order effect in respect with time.
These results significantly support our first hypothesis and allow us to assert that that population heterogeneity stems from audience heterogeneity and that an increase in parties heterogeneity is the response to changes and evolutions in audience s needs and tests. In respect with time variables, population heterogeneity is an inverted U-shaped function of time and doesn t reflects the tangentopoli effect. In fact we observe a constant increase in parties heterogeneity, while in most recent years prevails a bipolarization effect. The dummy variable doesn t have significance because the political system responded to its collapse with a high dynamicity and a total, more or less, replacement of traditional parties. This reaction didn t lead to an increase in the number of parties or in their heterogeneity, but it lead to a reconfiguration of the political scenario. In fact the bipolarization effect seems to be shared out on the last ten years. This is due to the incremental (not radical) process of transformation experienced by the political system: in 1992 corruption probes begun, in 1993 a referendum approved the moving from a proportional to a largely majoritarian electoral system, in February 1995 the majority reform passed and in 1999 the Parliament approved a constitutional law which states that the President of the regional Cabinet is elected by universal and direct suffrage. Virtually, the decrease of parties heterogeneity started in 1992-1993 but it s possible to comprehend the whole effect looking at the period 1993-2000. Model The full model specification used to analyze the founding rate ( λ (t) ) in the population of regional political parties in Region i is: 2 ( β N + β N + γ HET ) ( εed + φesi + ϕsc πtan) λ i ( t) = exp 1 i( t 1) 2 i( t 1) i( t 1) exp i( t 1) i( t 1) i( t 1) + (2) where λ (t) is the organizational founding rate of regional political parties population in Region i i and t = 1,2,,T identifies the time periods (years) for which data are available (1980-2003). The variables contained in equation (2) include: i
N i(t-1), the intra-population density of regional parties in Region i at time t-1, and its square N ; 2 i( t 1) HET i(t-1), the intra-population heterogeneity of regional parties in Region i at time t-1; ED i(t-1), the Economic Development heterogeneity of Region i at time t-1; ESI i(t-1), the Extension of Service Industry heterogeneity of Region i at time t-1; SC i(t-1), the Social Conflict heterogeneity of Region i at time t-1; TAN a dummy variable considering the Tangentopoli effect, coding 1 for years after 1993. Audience heterogeneity As the hypothesis suggests, a widening in the spectrum of organizations features belonging to a given population reflects an enlargement in audience heterogeneity, therefore we expect that an increase in socio-economical heterogeneity has a positive impact on founding rates ( ε > 0 ; φ > 0 ; ϕ > 0 ). Density We suppose the heterogeneity effect to be relevant and the density dependence model not to explain the births dynamics: we expect to find no significant correlation between density, and its square, and founding rates ( β = 1 0 and β = 0 ). 2 Heterogeneity We maintain that the key attribute of a population in describing births dynamics it s the heterogeneity within the population. Going into details, we assert that a disputable effect causes founding rates to rise with heterogeneity ( γ > 0 ). Methods of Analysis Following previous ecological research, organizational founding is conceptualized as an arrival process, the rate of which is affected (accelerated or decelerated) by institutional constraints related to regulation, by economic processes to relative size and market power of existing organizations, and finally, by ecological forces related to population density, changes in the level of resource availability, and community-level interactions between organizational populations. With notable
exception (Baum & Haveman, 1997) research on founding rates suffers from the limits to account for possible heterogeneity among founded organizations (Hannan & Freeman, 1989). Therefore a common approach it to treat the population as the unit experiencing the events. In this eventrecurrences framework organizational creation is conceptualized as a continuous-time discrete-state stochastic process of arrival into the population (Lomi, 1993; Hannan, 1991; Hannan & Freeman, 1989; Tuma & Hannan, 1984). The baseline model specification assumes that the number of new organizations appearing in the population at any given time period follows a Poisson process with arrival rate λ( t ), defined as: λ( t ) =lim Δt 0{Pr[Y(t + Δ t)-y(t) = 1 Y(t) = y]/ Δ t } where Y(t) is a non-negative random variable representing the cumulative number of arrivals in the population by time t. The Poisson regression model is defined in terms of its density Pr(Y=y t )=[exp(-λ)λ yt ]/ y t!, where y t is the observed number of events in period t, and E[y t ] = Var [y t ] = λ t. To test for the effects of observed covariates and period effects on the founding process, to compare the current analysis with previous organizational ecology research, and to ensure non-negativity of the birth rate, the empirical analysis of this study is based on models with general regression structure: λ( t ) = exp ( βx', θ ' z, θp ) t t t where x t is a vector of observable factors systematically affecting λ( t ) (containing a constant term) and considered in organizational ecology research, z t is a vector of general observable exogenous factors, and P t are different period effects. Previous research demonstrated that in empirical processes of organizational founding the Poisson assumption of a unitary coefficient of variation is often violated (Lomi, 1993). To account for overdispersion a common representation assumes the presence of a stochastic component in the rate and