Poor Institutions, Rich Mines: Resource Curse and the Origins of the Sicilian Mafia

Buonanno, Paolo; Durante, Ruben; Prarolo, Giovanni; Vanin, Paolo Working Paper Poor Institutions, Rich Mines: Resource Curse and the Origins of the Sicilian Mafia Quaderni - Working Paper DSE, No. 844 Provided in Cooperation with: University of Bologna, Department of Economics Suggested Citation: Buonanno, Paolo; Durante, Ruben; Prarolo, Giovanni; Vanin, Paolo (2012) : Poor Institutions, Rich Mines: Resource Curse and the Origins of the Sicilian Mafia, Quaderni - Working Paper DSE, No. 844, Alma Mater Studiorum - Università di Bologna, Dipartimento di Scienze Economiche (DSE), Bologna, https://doi.org/10.6092/unibo/amsacta/3999 This Version is available at: https://hdl.handle.net/10419/159683 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc/3.0/ Poor Institutions, Rich Mines: Resource Curse and the Origins of the Sicilian Mafia Paolo Buonanno Ruben Durante Giovanni Prarolo Paolo Vanin Quaderni - Working Paper DSE N° 844 Poor Institutions, Rich Mines: Resource Curse and the Origins of the Sicilian Mafia ∗ Paolo Buonanno University of Bergamo Ruben Durante Sciences Po Giovanni Prarolo University of Bologna Paolo Vanin University of Bologna First version: May, 2011 This version: August, 2012 Abstract This study explains the emergence of the Sicilian mafia in the XIX century as the product of the interaction between natural resource abundance and weak institutions. We advance the hypothesis that the mafia emerged after the collapse of the Bourbon Kingdom in a context characterized by a severe lack of state property-right enforcement in response to the rising demand for the protection of sulfur - Sicily’s most valuable export commodity - whose demand in the international markets was soaring at the time. We test this hypothesis combining data on the early presence of the mafia and on the distribution of sulfur reserves across Sicilian municipalities and find evidence of a positive and significant effect of sulphur availability on mafia’s diffusion. These results remain unchanged when including department fixed-effects and various geographical and historical controls, when controlling for spatial correlation, and when comparing pairs of neighboring municipalities with and without sulfur. Keywords: Natural Resource Curse, Weak Institutions, Mafia-type Organizations ∗Corresponding author: Giovanni Prarolo. Department of Economics, University of Bologna, Piazza Scaravilli 2, 40125 Bologna, Italy. [email protected]. We thank Giorgio Chiovelli, Francesco Cinnirella, Carl-Johan Dalgaard, Giovanni Federico, Oded Galor, Diego Gambetta, Paola Giuliano, Luigi Guiso, Nathan Nunn, Paolo Pinotti, Nancy Qian, participants at the Transatlantic Workshop on the Economics of Crime (Carlo Alberto, October 2011), the Frontier Research in Economic and Social History Meeting (Sant’Anna School of Advanced Studies, April 2012), the CEPREIEF Conference on Economics of Interactions and Culture (EIEF, April 2012) and the Conference on Intergenerational Transmission of Entrepreneurship, Occupations and Cultural Traits in the Process of Long-run Economic Growth (Naples Parthenope, May 2012) and seminar participants at DIW, Collegio Carlo Alberto, Universitat de Barcelona, University of Bologna, University of Neuchatel and Free University of Berlin, for valuable comments. 1 1 Introduction A large literature in economics and political science has investigated the effect of natural resources on political and economic development (Sachs and Warner, 1995, 2001; Mehlum et al., 2006a,b; Haber and Menaldo, 2011). These studies have delivered rather mixed results, and a general consensus has not emerged on whether, ultimately, resource abundance should be viewed as a “blessing” or as a “curse”.1But which factors explain why the discovery of valuable resources leads to desirable outcomes in some countries (e.g. Norway, Australia) and deleterious ones in others (e.g. Nigeria, Zimbabwe)? The quality of pre-existing political and legal institutions is arguably important: when institutions are dysfunctional, conflict over access to resource rents is likely to escalate, giving rise to increased corruption, rent-seeking, and even violence (Skaperdas, 2002; Collier and Hoeffler, 2002). Similarly, the literature on organized crime (Gambetta, 1993; Konrad and Skaperdas, 2012) has argued that the combination of weak institutions and resource abundance can be conducive to the emergence of mafia-type organizations which can have profound and long-lasting effect on a country’s economic prospects. The profound socio-economic consequences of organized crime has been a subject of growing interest among academics and policy-makers alike (Jennings, 1984; Fiorentini and Peltzman, 1997; Skaperdas, 2001). Research on the topic has focused, in particular, on the study of mafia-type organizations operating in various parts of the world.2While these contributions have expanded our knowledge of the nature and structure of such organizations, their economic origins remain largely unexplored. Our paper attempts to fill this gap advancing the hypothesis that mafias - which following Gambetta (1993) we conceptualize as providers of private protection - emerge to protect valuable natural resources when public law-enforcement institutions are weak or absent. While this argument is applicable to a broad range of examples, our empirical analysis focuses on the particular case of the Sicilian mafia, the oldest and most notorious example of this sort of organizations which dates back to the XIX century and which has had a considerable and long-lasting effect on Sicily’s socio-economic development.3More specifically, we argue that the Sicilian mafia emerged after the 1See Frankel (2010) for a comprehensive survey on the topic. 2Relevant contributions on mafia-type organizations by sociologists include Gambetta’s seminal work on the Sicilian mafia (1993), Varese’s studies of the Russian mafia (2005), and of the Japanese Yacuza. Recent contributions by economists have looked the economic impact of the Sicilian mafia (Pinotti, 2011) and the workings of criminal networks in the context of the American mafia (Mastrobuoni and Patacchini, 2011). 3Lupo (1993) and Dickie (2004) provide an excellent account of the history of the Sicilian mafia 2 demise of feudalism and the collapse of the Bourbon Kingdom in a context characterized by a severe lack of state property-right enforcement in response to the rising demand for the protection of sulfur - one of Sicily’s most valuable export commodity - whose demand in the international markets was soaring at the time. Such positive shock to the value of sulfur protection favored Mafia’s emergence in areas endowed with sulfur reserves, naturally more prone to its production and commercialization. To test this hypothesis we employ a comprehensive dataset which combines various measures of early mafia’s diffusion across Sicilian municipalities with detailed information on natural resource endowments and a range of other geographical characteristics. Our identification strategy exploits exogenous differences in the geographic distribution of sulfur reserves, which exposed ex-ante similar Sicilian municipalities to an asymmetric shock to the value of protection, which lasted for most of the XIX century. While documenting the magnitude of the shock and defending the exogeneity of its distribution is relatively straightforward, establishing that municipalities that experienced such a shock were ex-ante similar to those that did not is a more demanding task. One crucial difficulty lies in excluding that differences in the availability of natural resources are correlated to differences in other dimensions (e.g. institutional quality), which may affect mafia’s emergence through other channels. To address this concern, we pursue several strategies. First, we document that differences in sulfur reserves are not correlated with population growth rates in previous centuries. Since in the context of a Malthusian regime population represents a good indicator of the degree of economic development, this is indicative of the fact that sulfur played no special economic role prior to the XIX century. Second, in our econometric analysis we control for a wide range of observables that are likely to be correlated with institutional quality, economic activity and geographic and demographic differences. Third, in all our specifications we include area fixed effects, which allows us to identify our main effect from variations in sulfur endowment within small areas, which are plausibly homogeneous along several non-observable dimensions. Finally, following Acemoglu et al. (2012), we perform additional tests of our hypothesis based on the comparison of pairs of neighboring municipalities with different sulfur endowment. Furthermore, to account for possible spatial correlation in mafia’s emergence, we replicate our analysis using spatial regression methods.4 and of its expansion to other regions of Italy and to the United States. 4Acemoglu et al. (2012) employ this strategy to estimate the effect of gold-mines-related slavery in Colombia. One advantage of our application is that we do not have to worry about the endogeneity of slavery, since we are directly interested in the effect of natural resources. 3 Our empirical findings provide strong support for our main hypothesis. In particular, we find that sulfur availability has a positive large and significant effect on early incidence of mafia activities. Our findings are robust to the use of different measures of early mafia incidence, to the introduction of a number of geographical and socioeconomic controls, and to the use of the complementary approaches discussed above. We also discuss and test alternative explanations of the emergence of the Sicilian mafia proposed in the literature - such as the key role played by citrus production - which, however, do not appear to find support in the data. Although the focus of our analysis is on the emergence of the mafia, in the last part of the paper we also document the existence of a strong correlation between historical and current presence of the mafia, and discuss the possibility of using sulfur availability as an instrument for the latter. Although specific to the case of the Sicilian mafia, we believe that our findings can be helpful to inform our understanding of the rise of mafia-type organizations in various different parts of the world, where similar economic and institutional conditions may have occurred.5” The remainder of the paper is organized as follows. Section 2 illustrates our theoretical framework and relates our work to the literature on mafia-type organizations and resource curse. In section 3 we discuss the socio-economic and political conditions of XIX century Sicily that favored the emergence of the mafia, with particular regard for the collapse of the Bourbon regime and the upsurge in sulfur’s value. In section 4 we present the data used in the empirical analysis, while in section 5 we describe our empirical strategy and discuss our findings. Section 6 concludes. 2 On mafia and resource curse Various theoretical approaches have been proposed in the literature to study the structure and functioning of mafia-type organizations (see Fiorentini and Peltzman, 1997 and Anderson and Bandiera, 2005 among others). It is hence important to immediately clarify what is the working definition of mafia we refer to. Following Gambetta’s seminal contribution (1993), we conceptualize the mafia as an industry for private protection; in this framework the equilibrium level of mafia’s activity is determined by the interaction between demand and supply of protection services by private providers since public provision is lacking or greatly ineffective. 5It is the case, for example, of Yakuza in Japan, the Triad in Hong Kong and the Russian mafia. Indeed, Yakuza had its origins after the demise of the feudal system in Japan, while the Russian mafia after the dissolution of the USSR. 4 Our research relates to the vast literature on the socio-political impact of natural resources. This literature has discussed various mechanisms through which resource abundance may ultimately be regarded as a “curse”: vast resources may fuel violence, theft and looting (Skaperdas, 2002), they may be used to finance rebel groups, warlords or civil wars (Collier and Hoeffler, 2002), or may favor the emergence of criminal organizations aiming to extract part of the wealth derived from their exploitation. Although some evidence suggests that resource-rich countries display worse economic performance than resource-poor ones (Sachs and Warner, 1995, 2001), no unanimous consensus on this matter has emerged (see for example Haber and Menaldo, 2011). In fact, as some observers have pointed out, whether natural resources may result in a “curse” or a “blessing” may crucially depend on a country’s institutional quality (Mehlum et al., 2006a,b).6In the presence of weak institutions, our argument goes, natural resources are particularly vulnerable to predatory attacks; in this context, the (illegal) use of violence provides mafia-type criminal organizations with a competitive advantage in the supply of protection and extortion (Gambetta, 1993; Konrad and Skaperdas, 2012), resulting in the capacity to extract a substantial portion of naturalresource-based rents. This research integrates two previous econometric studies that have looked at the historical emergence of the Sicilian mafia. The first one, by Bandiera (2003), uses a common agency model to formalize the idea that the mafia should have been historically more active in towns were land was more fragmented,7and finds support for this hypothesis using qualitative data from the 1885 parliamentary survey (Damiani, 1885) on 70 districts (mandamenti) in western Sicily. 8The second one, by Pazzona (2010), expands Bandiera’s sample to 160 observations, documenting that the mafia was more likely to emerge where the competition by new social actors was harsher, particularly in areas where land value was higher and land holdings larger, at the opposite of what Bandiera (2003) finds. We improve upon these contributions by expanding the scope of the analysis to a much larger and more detailed set of geographical units covering the entire island; this allows us to investigate the large differences in the incidence of 6Institutional quality may, in turn, depends on features as diverse as geographic isolation, ethnic divisions, or state collapse (Skaperdas, 2011). 7The argument is based on the idea that the purchase of protection from a single landowner imposes a negative externality on the other ones (since it deflects thieves on their properties), and that, hence, landlords will be competing with each other to acquire protection and to exclude others from it. By increasing the number of competing landlords, land fragmentation should increase mafia’s potential profits. 8We present and discuss in detail in the data section the parliamentary survey employed by Bandiera (2003) and the administrative jurisdictions in XIX century Sicily. 5 the mafia across Sicilian areas, which is considered one of the most puzzling question about the history of the Sicilian mafia.9 Following the first version of this paper, two independent contributions have explored alternative explanations for the mergence of the Sicilian mafia. While Dimico et al. (2012) propose an argument similar to ours but centered around the historical role of citrus fruits, Del Monte and Pennacchio (2012) investigate the relationship between organized crime and brigandage. In our empirical section we test the robustness of our results to these alternative explanations and discuss some important data and methodological issues which, in our view, raise concern over the solidity of these studies’ respective findings. More in general, the results of this research complement the literature on the emergence of persistent social institutions as the consequence of what can be viewed as ‘historical accident’ (Acemoglu et al. (2001)), which in the context our our study, would be represented by the sudden rise in international demand for Sicilian sulfur. Although geographical characteristics per se are not the focus of our study, the findings we present can also be interpreted in the context of the debate on the long-term impact of geography on socio-economic development. Previous research has documented that the environment can influence economic performance directly, through its effect on health and agricultural productivity (Landes (1998); Sachs and Malaney (2002)), and indirectly, by setting the conditions in which social norms and political institutions have formed (Sokoloff and Engerman (2000); Easterly and Levine (2003); Durante (2009); Nunn and Puga (2012) and Michalopoulos et al. (2010)) or by defining environmental constraints to population growth (Galor and Weil (2000)). The evidence presented here suggests that, under given economic circumstances, geographic characteristics may have contributed to the emergence of particular forms of social organizations (criminal ones in this case), which have persisted over time and continue to have relevant socio-economic effects. 3 Historical background XIX century Sicily presented the two conditions that, according to the “resource curse” argument discussed above, are conducive to mafia’s emergence: poor quality of lawenforcement institutions and soaring value of domestic natural resources. In what follows we discuss some aspects of the XIX century Sicily political and socio-economic 9This aspect has been discussed by historians, sociologists and economists alike. Some examples include Lupo (1993); Gambetta (1993) and Sylos Labini (2003). 6 context that are relevant to our analysis, with particular regard to the main sources of institutional weakness, and the impact of growing international demand for Sicily’s high-value mining and agricultural production. 3.1 Institutional weakness and economic development Two major political transformations characterized the history of Sicily during the XIX century: the demise of feudalism in 1812, and the collapse of the Bourbon’s domination in 1861. Both these events contributed to the disruption of Sicilian law enforcement institutions and the deterioration of property-rights security. Feudal barons had long been struggling with the monarchy, which imposed on them a heavy fiscal burden, and this struggle intensified when, in 1806, in an attempt to escape from the French, the Bourbons moved from Naples to Palermo, Sicily’s capital. At the beginning of the XIX century most of Sicilian municipalities were under barons’ direct jurisdiction; however, most feudal lords did not reside in their lands but in Palermo, the center of the island’s political, economic and social life.10 Their lands, together with their feudal rights, were generally rented out to local administrators (gabelloti), who were in charge of managing the landholding’s productive activities and who invested their own capital in it. The abolition of feudalism represented a profound institutional change, which officially transferred all feudal jurisdictions to the State. Yet, while barons’ control over their lands had been weakened, limits to the power of the monarch had also been imposed by the establishment of a parliamentary system that assigned to the Parliament - largely dominated by the barons - extensive powers of control over the King’s acts (Candeloro, 1956). The power struggle between landlords and the Bourbons continued until 1861, when the kingdom collapsed and its territories were annexed to the newly formed Italian State. This institutional struggle resulted in extremely low levels of law enforcement, a situation which did not improve under the unified State. The feudal economy was primarily agrarian, based on extensive cultivations (mainly cereals) and characterized by very low productivity and peasants’s extreme poverty. The abolition of feudalism had little de facto impact on land distribution and did not result in increased productivity (Blok, 1966, 1969). The most valuable productions were particularly exposed to predatory attacks from local bandits (briganti), and the 10The port of Palermo was by far the more trafficked in the Island. In 1838, on a total of 480 Sicilian merchant vessels that left for foreign ports, 240 were from Palermo, while only 65 from Messina, the second port in order of importance (Petino, 1958). 7 experts (see Lupo, 1993 and Sylos Labini, 2003 among others), is further reassuring on the accuracy of Cutrera’s data.20 Figure 3: Geographic distribution of the Sicilian mafia in the late XIX century (a) Cutrera (b) Damiani-Jacini The figure reports the assessment of the intensity of mafia activity according to Cutrera (left) and Damiani-Jacini (right), with darker colors indicating more intense mafia’s activity (missing values are reported in white). Cutrera’s data are at the municipality level while Damiani-Jacini’s are at the district level. Missing values correspond to cases for which it was not possible to match historical municipalities (or districts) to current ones, e.g. for municipalities created in the XX century. 4.2 Sulfur and other geographical and historical controls With regard to the presence of sulfur, comprehensive municipal data are available from Squarzina (1963). These include information on the number of sulfur mines in each Sicilian municipality in 1886 - that is, around the pick of Sicily’s sulfur export boom. Since we are interested in gauging the original stock of sulphur available in each municipality - hence prior to the intense depletion which took place throughout the XIX century - we consider those mines that were still operating in 1886 as well as those that were already exhausted by then. It is worth emphasizing, once again, that Sicilian sulfur was generally superficial so that its extraction was relatively unchallenging and did not require considerable investments; as a consequence, at the peak of Sicily’s 20A visual inspection of the two maps highlights incoherent patterns in Damiani-Jacini’s data, that are at odds with a diffusive process as the mafia is likely to be. Take for example the northwestern area around Palermo: Palermo and its surrounding coastal municipalities are all characterized by high mafia activity, while most of the other neighboring southern municipalities are all coded as having no mafia activity. Going further south, again many municipalities turn to be flagged as high mafia ones. Conversely, Cutrera’s map shows a much smoother spatial pattern in terms of intensity of mafia activity, suggesting a more careful and homogeneous assessment of the variable. 14 sulfur export boom, virtually all major sulfur reserves on the island had been tapped (Squarzina, 1963). In light of this fact, the presence of sulfur mines can be considered a good proxy for the exogenous distribution of sulfur reserves, which is ultimately our variable of interest. The distribution of sulfur mines in each municipality (labeled sulfur henceforth), is summarized in Figure 4.2. Figure 4: Geographic distribution of Sicilian sulfur mines (1886) The figure reports the number of sulfur mines (both active and exhausted) recorded in each Sicilian municipality in 1886 on a four-color scale: 0 (lighter gray), 1-10, 11-30, more than 30 (darker gray). In our empirical analysis we control for a range of other geographical and historical variables at the municipal level. We focus, in particular, on factors that may have influenced the demand and supply of protection and, at the same time, may be correlated with the presence of sulfur, in order to test that sulphur availability has an independent effect on the emergence of the mafia and is not merely proxying for other characteristics. To account for difference in topography, we control for terrain ruggedness and elevation. As discussed by Nunn and Puga (2012), in addition to its obvious effect on agricultural productivity and trade, rugged terrain - in the form of hills, caves and cliffs - provides lookout posts and hiding places for individuals trying to escape. Arguably, in the context of Sicily rugged areas provided outlaws with better protection 15 from police forces. This view is consistent with accounts by various historians; for example, when discussing the widespread presence of the mafia in the mountainous towns of Gangi, Lupo (2004) emphasizes the importance of the town’s impervious location which made it particularly difficult for law enforcement officers to establish control over the surrounding area and apprehend criminals. To account for this aspect, we include in our regressions a municipal measure of terrain ruggedness constructed from the Global Land One-km Base Elevation Project (GLOBE), a global gridded digital elevation data set covering the Earth’s surface at a 10-minute spatial resolution (approximately 1km).21 Relatedly, we also control for difference in elevation within a given area, a variable which has been identified in the literature as imposing significant limitations on both agriculture and breeding activities (Michalopoulos, 2011; Grigg, 1995; Lupo, 1993). In particular, we use data on the maximum difference in altitude in a given municipality available from the Italian Institute of Statistics (ISTAT). We are also interested in controlling for the suitability of local land for various types of crops, as this is likely to have affected the demand for protection of agricultural goods. In fact, while some low-value crops (i.e. cereals) were consumed locally, others (e.g. citrus fruits, olive, sumac) were high-value export goods which demanded a degree of protection similar to that of sulfur. To account for this aspect, we include in our regressions measures of land suitability for the three most widespread crop categories in XIX century Sicily: i) citrus fruits, ii) cereals, and iii) olives. These measures are constructed using data on crop-specific agro-ecological suitability available from the IIASA-FAO Global Agro-Ecological Zones project (GAEZ).22 The GAEZ data are in grid format, have a very high resolution (1’), and assign to each grid cell a value from 0 (totally unsuitable), to 100 (very suitable). To obtain suitability measures at the municipality level we average the individual suitability score of all the cells in a given municipality. We focus on crop suitability - determined in large part by exogenous soil properties and climatic conditions - rather than on actual crop production to minimize concerns of possible reverse effects of the presence of the mafia on the prevalence of particular crops. Nevertheless, in the last part of our analysis, to test one of the alternative hypothesis about the emergence of the mafia, we use information on the proportion of land devoted to the cultivation of citrus fruits in each municipality, 21The GLOBE data set has superseded the GTOP30 which, before the introduction of GLOBE, was considered the most accurate digital elevation data set and had been used, among others, by Nunn and Puga (2012). 22More information on the FAO-GAEZ project can be found at http://www.gaez.iiasa.ac.at/. Data from FAO-GAEZ have been used extensively by economists in recent years to investigate a variety of topics. Examples include Nunn and Qian (2011), Michalopoulos (2011) and Durante (2009). 16 available from the Damiani-Jacini’s inquiry. Given the importance of irrigation for several of the crops mentioned above, and in light of the accounts of various observers about the crucial role played by the mafia in controlling water supply, we also control for the relative scarcity of water in a given area using data on the presence of underground water basins available from the Sicilian Waters Observatory. Another set of controls is intended to capture factors such as access to major ways of communication and proximity to the main ports, which were arguably important determinants of the value of protection. The first variable, labeled as postal roads, indicates whether, at the beginning of the XIX century, a municipality had direct access to one of the postal roads which connected Sicily’s largest towns. The data are derived from the digitalization and geo-referentiation of a detailed historical map of Sicily in late XVIII century (Cary, 1799), hence prior to the steady rise in sulfur’s international demand. Other variables include the distance of a municipality’s centroid from the closest non-seasonal river (river distance), and distance from the closest commercial port (port distance).23 We also control for a set of socio-economic and demographic characteristics which could potentially be related to both sulphur availability and mafia presence. In particular, to account for the fact that crime incidence might be higher in more densely populated areas (Glaeser et al., 1996; Glaeser and Sacerdote, 1999; Buonanno et al., 2012) we control for a measure of population density (density) based on data on municipal population from the 1861 census.24 Relatedly, to control for possible differences between rural and urban areas, we also define a dummy variable, urban, indicating whether a municipality is located at a distance of 10Km or less from one of Sicily’s then five largest cities.25 Finally, we also control for the degree of land fragmentation, a factor which previous contributions have related to the development of a florid market for private protection and the consequent emergence of the mafia (Bandiera, 2003). Information on the degree of land fragmentation in each municipality is available from one of the questionnaires 23Sicily’s main commercial ports were: Mazara del Vallo, Porto Empedocle, Trapani, Siracusa, Catania, Palermo and Messina. 24By 1861 Sicily’s total population amounted to 2.1 million, accounting for more than 10% of Italy’s population. Population density was more than 81 inhabitants per squared kilometer, roughly comparable to the current density of Spain. Since then, the increase in population has been rather homogeneous across Sicilian municipalities, resulting in a correlation between population in 2001 and in 1861of of 0.95. Palermo, the capital, was Sicily’s largest and denser city, with a population of 185,000 inhabitants and a density of 1,000 inhabitants per squared kilometer, comparable to that of current mid-size Italian cities. 25These include: Palermo, Catania, Trapani, Messina and Girgenti. 17 of the Damiani-Jacini’s inquiry in which mayors where asked to report whether land in their municipality was prevalently composed by small, medium or large landholdings. In particular, we define a dummy variable, fragmentation, taking value one for municipality where small and medium landholdings were prevalent, and zero in those in which large landholding still existed. Due to the rather low response rate to this question, data on fragmentation are available for only 237 out of the 285 municipalities in our sample. The availability of data at the municipal level allow us to include in all our regression department specific fixed effects which capture the political and historical background common to municipalities in the same department. In particular, since in the period under examination the Sicilian administrative, judicial and law-enforcement systems were organized at department level, the inclusion of 24 department fixed effects allow us to estimate the effect of sulphur on mafia by comparing municipalities characterized by a fairly homogenous level of institutional quality. 5 Empirical Analysis This section presents the empirical assessment of how geographical variations in sulfur endowment contributed to the emergence of the Sicilian mafia. As discussed above, over the XIX century Sicily experienced the collapse of the Bourbon Kingdom and a generalized situation of weak law enforcement. Over the same century, international demand for sulfur soared and most of the world supply came from Sicily. Municipalities with sulfur reserves thus experienced a boom in the value of their natural resources.26 We exploit the exogenous distribution of sulfur reserves to identify the effects of such boom on mafia’s emergence. We document a resource curse, by which valuable natural resources fostered protection demand and extortion opportunities, thus favoring the emergence of organized crime. 5.1 Municipality-level estimates Table 2 provides our first clear evidence of the importance of sulfur for mafia’s emergence. It reports municipality-level OLS estimates of our preferred mafia measure 26As already shown in figure 1, sulfur export in Sicily grew at an impressive rate of 9% throughout the period 1830 and 1860 and in that period Sicily served around 90% of the world sulfur demand. sulfur export was negligible at the beginning of the XIX century, peaked at the end of that century (reaching 540,000 tons in 1901) and sharply declined in the XX century (by 1976, it was only 85,000 tons). 18 (maf c) on the number of sulfur mines (sulfur). The different columns gradually increase the number of control variables.27 Column 1 shows that, in a univariate regression (including a constant, as in all regressions in all tables), the estimated coefficient on sulfur, significant at the 1% level, is equal to 0.033, implying that a one standard deviation increase in sulfur leads to an increase in maf c by more than one sixth of a standard deviation.28 Column 2 adds department fixed effects. This is our first step in tackling the issue of whether differences in sulfur endowment pick up differences in other variables, which may matter for mafia’s emergence. Such fixed effects control for any characteristic that was common within each department. The result shows that even within each department, municipality-level variations in sulfur endowment were positively and significantly associated to variations in early mafia’s presence (the point estimate is 0.022 and it is significant at the 1% level). To minimize the risk that within-department variations in sulfur endowment are related to differences in other variables, which may themselves be related to mafia activity, columns 3 to 7 progressively add municipality-level controls for differences in agriculture, geography, transportation and communication, socio-demographic variables and land fragmentation (always including department fixed effects). No matter which controls we include, sulfur remains highly significant and the magnitude of the coefficient is barely affected, suggesting that our main result is not driven by any omitted variable.29 The first group of additional control variables contains exogenous soil characteristics, which are related to agricultural activity: column 3 includes soil suitability for cultivation of citrus fruits, cereals and olives, as well as a dummy for water scarcity. It is important to control for such factors because many scholars have argued that mafia’s emergence was related to citrus cultivation (Lupo, 2004; Gambetta, 1993; Dickie, 2004; Del Monte and Pennacchio, 2012; Dimico et al., 2012) and also to mafia’s possibility to control scarce water resources and thus have high extortionary power towards agricultural production (Sylos Labini, 2003). Both suitability for cereals and for citrus have a significant effect, the former positive and the latter negative, but only the former effect remains significant as additional controls are introduced. These results do not support 27Since maf c is an ordinal variable, we repeated the entire analysis using ordered probit, obtaining analogous results, which are available upon request. 28Given the linearity implied by the measure of mafia activity, it takes 30 mines more to have an increase in mafia activity from one level (not present, presence is low, intermediate, high) to the next. 29The coefficient on sulfur is only raised when we add land fragmentation to the controls. Yet, notice that this reduces sample size from 282 to 237 municipalities since, as discussed in the previous section, information on fragmentation are available only for 237 municipalities. 19 the widely held idea that early mafia’s development was related to citrus cultivation. Column 4 adds two geographic controls: average ruggedness and difference in elevation (Nunn and Puga, 2012; Michalopoulos, 2011). Ruggedness is never significant, while difference in elevation is strongly significant and positive. Column 5 adds the presence of 1799 postal roads, as well as distance from the closest non-seasonal river and from the closest commercial port. Mafia’s presence was significantly higher in municipalities along old postal roads, whereas distances from rivers and ports are not significant. Column 6 further adds population density in 1861 and a dummy for urban municipalities. Such variables are related to the level of economic activity, both legal and illegal (see, e.g. Glaeser and Sacerdote, 1999; Buonanno et al., 2012). In line with the literature, we find that population density is positively and significantly related to mafia’s emergence. Finally, column 7 adds land fragmentation, which according to Bandiera (2003) should be relevant for mafia’s emergence, but we find no support for her thesis. As an additional robustness check, we re-run our baseline regression substituting the number of sulfur mines (sulfur) with a dummy (sulfur d) for the presence of at least a mine in the municipality. Using sulfur d is important because we do not have information on the quantity of sulfur extracted in each mine and we therefore put ourselves in the worst case scenario, that is we do not exploit at all the information on the intensive margin in the distribution of sulfur. Table 3 reproduces the same specifications of Table 2, but it replaces sulfur d (the dummy for the presence of at least a sulfur mine in a municipality) for sulfur. For the sake of space, we only report the coefficient of sulfur d. Results show that, even if we neglect the information on the intensive margin of sulfur extraction and only rely on the extensive margin, the presence of sulfur is still significant to explain Mafia’s emergence. Yet, the slight decrease in significance levels also suggests that the intensive margin was relevant as well. While the above analysis suggests that omitted variables are not driving our results, as an additional check we control for possible spatial effects. There is no reason to believe that mafia’s activity follows the administrative boundaries of municipalities. Mafia lords may indeed offer protection and practice extortion in neighboring municipalities, whose territory they control. They may also establish agreements with other mafia lords, who control different territories, for instance to grant protection to their clients’ goods transiting through them. Moreover, protection externalities may foster mafia’s activity in a municipality just because the mafia is active in neighboring municipalities. There may therefore be relevant spatial spillovers from a municipality to its neighbors. Omitting to take them into account may reduce the efficiency of our estimates and bias them. 20 To address this issue, we estimate a spatial model by means of the generalized spatial two stage least squares (GS2SLS) estimator of Kelejian and Prucha (1998). Results are presented in table 4, which reproduces the same specification of column 6 of table 2.30 We employ both a non-standardized (columns 1 to 3) and a row-standardized (columns 4 to 6) contiguity matrix. We implement a spatial error model (columns 1 and 4), a spatial autoregressive model (columns 2 and 5) and a model that combines the two by considering both a spatial lag and a spatial error structure (columns 3 and 6).31 Spatial analysis is consistent with our baseline estimates. Interestingly, the coefficients on the spatial structure are almost always significant when using the rowstandardized contiguity matrix, suggesting that mafia’s activity in neighboring municipalities is indeed relevant and thus spatial estimates are justified. Yet, the sign of the spatial lag is not robust across specifications, so one should be cautious in its interpretation. By contrast, the main result of the spatial analysis is that, across all specifications, sulfur is always positively and significantly related to mafia density, with a point estimated which is very close to the one estimated in table 2, column 6. The above analysis controls for a great number of observed variables, as well as for any unobserved variable that is common to all municipalities in a department. Yet, in principle there is still the possibility that sulfur-rich municipalities differ from other municipalities of the same department along some other unobserved characteristic, that might be relevant for mafia’s emergence. This for instance might be the case if sulfur had always been a relevant source of rents, even in previous centuries. Although we have already documented that sulfur exports were negligible at the beginning of the XIX century, to indirectly test for this possibility we investigate whether sulfur presence was correlated to population growth in the XVII and XVIII centuries.32 We 30We use column 6 rather than column 7 to avoid the reduction in sample size implied by missing data for land fragmentation. Yet, results are robust to the inclusion of the latter variable. 31If neighboring units have similar intercepts due to their proximity, spatial dependence appears only in the error term (LeSage and Pace, 2009) and a Spatial Error model (SEM) should be estimated. In that case, omitting the spatial specification of the error term would reduce efficiency of the estimator, while preserving consistency (Anselin, 1988). In turn, if mafia’s density in one municipality is directly affected by mafia’s activity in neighboring locations, one should estimate a Spatial Autoregressive model (SAR), which includes among regressors a spatial lag, that is, a weighted average of mafia’s activity in neighboring municipalities. A (non-standardized) contiguity is a proximity matrix that associates 1 to each pair of municipalities sharing a border and 0 to any other pair (the diagonal is set to 0 by convention). Row-standardization is obtained by normalizing the sum of each row of the matrix to 1. The difference between the first and the second case is that the spatial lag captures total and average mafia’s activity in neighboring municipalities, respectively. 32Under the assumption that Sicily was on a Malthusian development path before the XIX century, if sulfur was a relevant source of rents, this should translate in higher rates of population growth where it was present. 21 focus on those Sicilian municipalities for which we have data for years between 1600 and 1800 (a period characterized by Malthusian regime in which development is proxied by population growth) and construct long term yearly population growth rates.33 Different specifications of a growth regression that have sulfur (and log of initial population to control for convergence) as controls show that sulfur had no economic role in the economic development of Sicilian municipalities before the beginning of the XIX century. Results are reported in table 5. 5.2 District-level estimates In our baseline regressions we used our preferred measure of mafia intensity, maf c, which is available at the municipality level. In order to obtain estimates directly comparable with other studies (Bandiera, 2003; Pazzona, 2010; Dimico et al., 2012), we employ the mafia measure as defined in the Damiani-Jacini parliamentary inquiry, maf d, which is only available at district level, for 158 Sicilian districts. We replicate the analysis proposed in our baseline regression presented in table 2, with the same specification: regressing maf d on sulfur and the controls presented and discussed in the previous sections. Since estimates exploit district-level information on mafia’s activity, we correspondingly re-define all our regressors at this level of geographical and administrative aggregation. District-level findings are presented in table 6. Throughout all the regressions, the estimated coefficient on sulfur is strongly significant and is extremely stable, suggesting that our results are not driven by any omitted variable. As previously stated, the use of maf d is not only useful as a robustness check, but it also allows a more direct comparison with earlier and subsequent contributions. In particular, Bandiera (2003), who also uses district-level data, provides early evidence, based on 70 districts located in the western part of Sicily, supporting the idea that land fragmentation may have favored mafia’s emergence; while on the contrary Pazzona (2010) provides evidence that the origins of the Sicilian mafia are rooted in the presence of large landholdings. Our results, both those based on all the 158 available districts, as well as those presented in Table 2, based on 237 municipalities, do not support their arguments. 33Data for population of Italian towns between 1300 and 1861 are available from Paolo Malanima at: http://www.paolomalanima.it/DEFAULT files/Page646.htm 22 5.3 Neighbor-pair fixed effects So far we have presented consistent and robust findings, documenting the significant effect of the presence of sulfur on mafia’s origins. Since sulfur is not randomly distributed across Sicily, but rather geographically concentrated, we have relied on department fixed effects and on municipality-level controls to make sure that differences in sulfur endowment do not pick up the effects of some other characteristics, which may be relevant for mafia’s emergence. In this section we go even deeper and, rather than comparing municipalities with different sulfur endowments within a given department, we follow Acemoglu et al. (2012) and exploit variations in sulfur endowments across direct neighbors. In particular, we restrict our analysis to the 48 municipalities which have sulfur mines and the 54 municipalities without sulfur mines which are adjacent to them.34 As in Acemoglu et al. (2012), we implement the neighbor-pair fixed effects estimator, very similar to a matching methodology and to a regression discontinuity design, comparing each sulfur-mining municipality to each of its neighbors. Figure 5 visually presents municipalities with sulfur mines and their neighbors. This empirical strategy makes it possible to directly control for unobservables that are common across adjacent municipalities by including neighbor-pair fixed effects. Indeed, we rely on the assumption that adjacent municipalities faced similar institutional and contextual conditions (i.e. law enforcement, state presence, culture, labor market, geography), and are likely to be very similar across any other unobservables. Within the neighbor pair, we claim that the exogenous source of variation in mafia’s activity is the presence of sulfur mines. Formally (see Acemoglu et al. (2012) for a more complete description), we define with Sthe subset of municipalities with sulfur mines and with N(s) all the adjacent municipalities without sulfur mines of each element of S. We use sand ito index municipalities with and without sulfur mines, respectively. We estimate the following model by means of OLS: maf cs=βsulfurs+γX0 s+ψsi +νss∈S(1) maf ci=βsulfuri+γX0 i+ψsi +νii∈N(s) (2) where X0 tcollects municipality-level controls, ψsi represents common unobservables for 34Note that although 48 municipalities have sulfur mines, 13 of them have as only neighbors other municipalities with sulfur, so they cannot be exploited in this analysis. 23 Glaeser, Edward L. and Bruce Sacerdote, “Why Is There More Crime in Cities?,” Journal of Political Economy, December 1999, 107 (S6), S225–S258. , , and Jos´e A. Scheinkman, “Crime and Social Interactions,” The Quarterly Journal of Economics, May 1996, 111 (2), 507–48. Grigg, D.B.,An introduction to agricultural geography, Routledge, 1995. Haber, Stephen and Victor Menaldo, “Do Natural Resources Fuel Authoritarianism? A Reappraisal of the Resource Curse,” American Political Science Review, 2011, 105 (01), 1–26. Jennings, William P., “A note on the economics of organized crime,” Eastern Economic Journal, 1984, 10 (3), 315–321. Kelejian, Harry H. and Ingmar R. Prucha, “A Generalized Spatial Two–Stage Least Squares Procedure for Estimating a Spatial Autoregressive Model with Autoregressive Disturbances,” Journal of Real Estate Finance and Economics, 1998, 17 (1), 99–121. Konrad, Kai and Stergios Skaperdas, “The market for protection and the origin of the state,” Economic Theory, 2012, 50, 417–443. Landes, D.S., “The wealth and poverty of nations: Why some are so rich and some so poor,” New York, 1998. LeSage, James and R. Kelley Pace,Introduction to Spatial Econometrics, London: CRC Press, Taylor & Francis Group, 2009. Lupo, Salvatore, “Storia della mafia dalle origini ai nostri giorni,” Donzelli, Roma, 1993. ,Storia della Mafia dalle origini ai giorni nostri, Donzelli, Roma, 2004. Mastrobuoni, Giovanni and Eleonora Patacchini, “Organized Crime Networks: an Application of Network Analysis Techniques to the American Mafia,” 2011. Mehlum, Halvor, Karl Moene, and Ragnar Torvik, “Cursed by Resources or Institutions?,” The World Economy, 2006, 29 (8), 1117–1131. , , and , “Institutions and the Resource Curse,” The Economic Journal, 2006, 116 (508), 1–20. 30 Michalopoulos, Stelios, “The Origins of Ethnolinguistic Diversity,” American Economic Review, 2011, forthcoming. , Alireza Naghavi, and Giovanni Prarolo, “Trade and geography in the economic origins of Islam: theory and evidence,” Fondazione Eni Enrico Mattei Working Papers, 2010, p. 464. Nunn, Nathan and Diego Puga, “Ruggedness: The Blessing of Bad Geography in Africa,” The Review of Economics and Statistics, 2012, 94 (1), 20–36. and Nancy Qian, “The Potato’s Contribution to Population and Urbanization: Evidence From A Historical Experiment,” The Quarterly Journal of Economics, 2011, 126 (2), 593–650. Pazzona, Matteo, “Emerging Classes and the Fight for Resources: an Empirical Investigation into the Origins of the Sicilian Mafia,” 2010. Pescosolido, Guido, “L’economia siciliana nell’unificazione italiana,” Mediterranea, 2010, 19, 217–234. Petino, Antonio,I prezzi di alcuni prodotti agricoli sui mercati di Palermo e Catania dal 1801 al 1890, Archivio Economico dell’Unificazione, Roma, 1958. Pinotti, Paolo, “The Economic Consequences of Organized Crime: Evidence from Southern Italy,” 2011. Rawson, W. Rawson, “On the Sulphur Trade of Sicily, and the Commercial Relations between that Country and Great Britain,” Journal of the Statistical Society of London, 1840, 2(6), 446–457. Sachs, J. and P. Malaney, “The economic and social burden of malaria,” Nature, 2002, 415 (6872), 680–685. Sachs, Jeffrey D. and Andrew M. Warner, “Natural Resource Abundance and Economic Growth,” Working Paper 5398, National Bureau of Economic Research 1995. and , “The curse of natural resources,” European Economic Review, 2001, 45 (4-6), 827–838. Skaperdas, Stergios, “The political economy of organized crime: providing protection when the state does not,” Economics of Governance, 2001, 2(3), 173–202. 31 , “Warlord Competition,” Journal of Peace Research, 2002, 39 (4), 435–446. , “The costs of organized violence: a review of the evidence,” Economics of Governance, 2011, 12 (1), 1–23. Sokoloff, Kenneth L. and Stanley L. Engerman, “History Lessons: Institutions, Factors Endowments, and Paths of Development in the New World,” The Journal of Economic Perspectives, 2000, pp. 217–232. Squarzina, Federico,Produzione e Commercio dello Zolfo in Sicilia nel Secolo XIX, Industria Libraria Tipografica Editrice, Torino, 1963. Sylos Labini, Paolo, “Le radici della Mafia in Sicilia,” in P. Sylos Labini and G. Arena, eds., Scritti sul mezzogiorno (1954-2001), Piero Lacaita Editore, Manduria-Bari-Roma, 2003. Varese, Federico,The Russian Mafia: private protection in a new market economy, Oxford University Press, USA, 2005. 32 Table 1: Descriptive Statistics Variable Obs Mean Std.Dev. Min Max maf c 282 1.433 1.140 0 3 maf d 158 .689 1.064 0 3 Sulfur 282 1.986 7.099 0 61 Sulfur dummy 282 .167 .373 0 1 Citrus suitability 282 15.608 7.658 0 48 Cereals suitability 282 17.728 11.149 1.490 66.380 Olive suitability 282 30.906 12.065 3.478 69.273 Water scarcity 282 0.702 0.458 0 1 Ruggedness 282 433.630 195.940 58.017 1,149.332 Diff. elevation 282 796.837 519.126 48 3,232 Postal roads 282 0.550 0.498 0 1 River distance 282 9.279 7.247 .992 42.075 Port distance 282 37.924 19.371 0.132 83.919 Urban 282 0.124 0.330 0 1 Density 282 132.412 126.861 4.856 1,177.986 Fragmentation 237 0.759 0.428 0 1 Note: Descriptive statistics of the main variables used in the empirical analysis. Data is at the municipality level except for maf d that is collected at the district level. 33 Table 2: Baseline estimates Dependent variable: maf c (1) (2) (3) (4) (5) (6) (7) Sulfur .0329∗∗∗ .0224∗∗∗ .0237∗∗∗ .0228∗∗ .0234∗∗ .0239∗∗∗ .0549∗∗∗ (.0105) (.0082) (.0089) (.0091) (.0092) (.0089) (.0127) Citrus suitabiliy -.0251∗-.0155 -.0235 -.0227 -.0293 (.0150) (.0151) (.0163) (.0166) (.0180) Cereals suitabiliy .0235∗∗ .0223∗∗ .0224∗∗ .0224∗∗ .0296∗∗∗ (.0108) (.0108) (.0107) (.0110) (.0111) Olive suitabiliy -.0037 -.0039 .0009 -.0016 -.0001 (.0119) (.0119) (.0131) (.0129) (.0145) Water scarcity .1267 -.0460 -.0231 .0057 -.0548 (.1986) (.1990) (.1923) (.1933) (.2010) Ruggedness -.0011 -.0010 -.0011 -.0006 (.0008) (.0008) (.0008) (.0009) Diff. elevation .0004∗∗∗ .0004∗∗∗ .0005∗∗∗ .0006∗∗∗ (.0001) (.0001) (.0001) (.0002) Postal roads .0811 .0975 .1470 (.1008) (.1002) (.1127) River distance .0101 .0045 .0091 (.0092) (.0094) (.0101) Port distance -.0087 -.0034 -.0022 (.0060) (.0069) (.0080) Urban .1855 .0925 (.1898) (.2133) Density .0013∗∗∗ .0012∗∗ (.0005) (.0005) Fragmentation .0961 (.1315) Department FEs N Y Y Y Y Y Y Obs. 282 282 282 282 282 282 237 R20.042 0.567 0.577 0.594 0.601 0.618 0.659 Note: This table presents the results of OLS estimates for Sicilian municipalities for which values for all the variables are available. The dependent variable is maf c, the level of mafia activity at the end of XIX century as coded by Cutrera (1900) on a 0 to 3 scale (0 is no mafia activity, 3 is large mafia activity). The main explanatory variable Sulfur is the number of sulfur mines as collected by Squarzina (1963), while the other control variables are described in the main text. Department fixed effects are included in all specifications except the first. Robust standard errors are presented in parentheses. *, ** and *** denote rejection of the null hypothesis of the coefficient being equal to 0 at 10%, 5% and 1% significance level, respectively. 34 Table 3: Baseline with sulfur dummy Dependent variable: maf c (1) (2) (3) (4) (5) (6) (7) Sulfur dummy .5532∗∗∗ .3594∗.4501∗∗ .3968∗.4165∗.4916∗∗ .4550∗∗ (.1765) (.2193) (.2286) (.2370) (.2449) (.2237) (.2168) Controls Department FEs N Y Y Y Y Y Y Suitability and water N N Y Y Y Y Y Geomorphological N N N Y Y Y Y Distances N N N N Y Y Y Sociodemographic N N N N N Y Y Obs. 282 282 282 282 282 282 237 R20.033 0.559 0.570 0.587 0.593 0.613 0.639 Note: Note: This table presents the results of OLS estimates for Sicilian municipalities, for which values for all the variables are available. The dependent variable is maf c, the level of mafia activity at the end of XIX century as coded by Cutrera (1900) on a 0 to 3 scale (0 is no mafia activity, 3 is large mafia activity). The main explanatory variable, Sulfur dummy, is a dummy taking value one if the number of sulfur mines as collected by Squarzina (1963) is greater than zero, while the other control variables are described in the main text. Department fixed effects are included in all specifications except the first. Robust standard errors are presented in parentheses. *, ** and *** denote rejection of the null hypothesis of the coefficient being equal to 0 at 10%, 5% and 1% significance level, respectively. 35 Table 4: Spatial estimates Dependent variable: maf c (1) (2) (3) (4) (5) (6) Sulfur .0229∗∗∗ .0234∗∗∗ .0227∗∗∗ .0211∗∗∗ .0234∗∗∗ .0231∗∗∗ (.0072) (.0071) (.0072) (.0071) (.0071) (.0066) λ.0143 .0149 .4762∗∗∗ .6857∗∗∗ (.010) (.010) (.138) (.113) ρ.0144 -.0038 0.0497 -0.6252∗∗∗ (.022) (.025) (.103) (.149) Controls Department FEs N Y Y Y Y Y Suitability and water N N Y Y Y Y Geomorphological N N N Y Y Y Distances N N N N Y Y Sociodemographic N N N N N Y Obs. 282 282 282 282 282 282 Note: This table presents the results of a spatial model estimated by means of the generalized spatial two stage least squares (GS2SLS) estimator of Kelejian and Prucha (1998). Included controls are the same as in the specification of column 6 of table 2. Columns 1 to 3 employ a non-standardized contiguity matrix, while a rowstandardized one is used in columns 4 to 6. A Spatial Error model, a Spatial Autoregressive model and a model that combines the two by considering both a spatial lag and a spatial error structure are respectively presented in columns 1 and 4, columns 2 and 5 and columns 3 and 6. λis the spatial error term, while ρis the spatial lag. The dependent variable is maf c, the level of mafia activity at the end of XIX century as coded by Cutrera (1900) on a 0 to 3 scale (0 is no mafia activity, 3 is large mafia activity). The main explanatory variable, Sulfur, is the number of sulfur mines as collected by Squarzina (1963), while the other control variables are described in the main text. Department fixed effects are included in all specifications except the first. Robust standard errors are presented in parentheses. *, ** and *** denote rejection of the null hypothesis of the coefficient being equal to 0 at 10%, 5% and 1% significance level, respectively. 36 Table 5: Growth regressions Dependent variable: population growth 1600-1700 1700-1800 1600-1800 (1) (2) (3) ln(population 1600) -.1660∗∗∗ -.0900∗∗ (.0576) (.0404) ln(population 1700) -.0761 (.0608) Sulfur .0003 -.0002 -.0002 (.0030) (.0028) (.0022) Obs. 47 56 50 R2.159 .029 .097 Note: This table presents the results of OLS estimates for Sicilian municipalities for which population was positive in at least two of the years 1600, 1700 or 1800, according to Malanima’s data (http://www.paolomalanima.it/DEFAULT files/Page646.htm). The dependent variable is the yearly population growth while the explanatory variables are the log of population at the beginning of the period and Sulfur, the number of sulfur mines as collected by Squarzina (1963). Robust standard errors are presented in parentheses. *, ** and *** denote rejection of the null hypothesis of the coefficient being equal to 0 at 10%, 5% and 1% significance level, respectively. 37 Table 6: District level estimates Dependent variable: maf d (1) (2) (3) (4) (5) (6) (7) Sulfur .0354∗∗∗ .0318∗∗∗ .0325∗∗∗ .0327∗∗∗ .0326∗∗∗ .0347∗∗∗ .0322∗∗ (.0080) (.0115) (.0116) (.0117) (.0121) (.0122) (.0127) Citrus suitability .0413 .0483 .0444 .0399 .0378 (.0315) (.0324) (.0344) (.0351) (.0352) Cereals suitability -.0034 .0024 .0022 .0042 .0016 (.0193) (.0188) (.0196) (.0197) (.0201) Olive suitability -.0203 -.0203 -.0212 -.0207 -.0155 (.0180) (.0177) (.0198) (.0199) (.0210) Water scarcity .2763∗.2842 .2669 .2365 .2078 (.1678) (.1730) (.1711) (.1720) (.1820) Ruggedness .0010 .0013 .0009 .0007 (.0002) (0.0002) (.0002) (.0002) Diff. elevation .0002 .0002 .0003 .0003 (.0002) (.0002) (.0002) (.0002) Postal roads .2210 .2877 .2680 (.1906) (.1947) (.1993) River distance .0026 -.0022 -.0040 (.0177) (.0182) (.0184) Port distance -.0009 .0016 .0002 (.0126) (.0145) (.0144) Urban -.1216 -.1115 (.4075) (.4097) Density .0014∗.0014∗ (.0008) (.0008) Fragmentation -.1847 (.2467) Department FEs N Y Y Y Y Y Y Obs. 158 158 158 158 158 158 158 R20.105 0.232 0.249 0.259 0.266 0.283 0.288 Note: This table presents the results of OLS estimates for Sicilian districts for which values for all the variables are included. The dependent variable is maf d, the level of mafia activity around 1883 as coded by Damiani (1885) on a 0 to 3 scale (0 is no mafia activity, 3 is large mafia activity). The main explanatory variable Sulfur is the number of sulfur mines as collected by Squarzina (1963), while the other control variables are described in the main text. Department fixed effects are included in all specifications except the first. Robust standard errors are presented in parentheses. *, ** and *** denote rejection of the null hypothesis of the coefficient being equal to 0 at 10%, 5% and 1% significance level, respectively. 38 Table 7: Neighbor pair fixed effect estimates Dependent variable: maf c (1) (2) (3) (4) (5) (6) Sulfur .0206∗∗∗ .0172∗∗ .0231∗∗ .0216∗∗ .0196∗∗ .0207∗∗∗ (.0061) (.0073) (.0093) (.0094) (.0094) (.0055) Controls Department FEs N Y Y Y Y Y Suitability and water N N Y Y Y Y Geomorphological N N N Y Y Y Distances N N N N Y Y Sociodemographic N N N N N Y Obs. 162 162 162 162 162 162 R20.054 0.637 0.667 0.672 0.681 0.795 Note: This table presents the results of OLS estimates for Sicilian municipalities for which values for all the variables are available. Observations are all those municipalities that form a couple in which a municipality has sulfur and its neighbor has not. Each municipality in a pair shares a common pair fixed effect. The dependent variable is maf c, the level of mafia activity at the end of XIX century as coded by Cutrera (1900) on a 0 to 3 scale (0 is no mafia activity, 3 is large mafia activity). The main explanatory variable Sulfur is the number of sulfur mines as collected by Squarzina (1963), while the other control variables are described in the main text. Pair fixed effects are included in all specifications. Robust standard errors are presented in parentheses. *, ** and *** denote rejection of the null hypothesis of the coefficient being equal to 0 at 10%, 5% and 1% significance level, respectively. 39