Organized Crime Networks: Theory and Evidence Based on Federal Bureau of Narcotics Secret Files on American Mafia preliminary version, please do not circulate ∗ Giovanni Mastrobuoni † June 2011‡
∗
I would like to thank Theo Diasakos, Matthew Jackson, Claudio Lucifora, Franco Peracchi, Jesse Rothstein, Rocco Sciarrone, Serena Uccello, Aleksey Tetenov and seminar participants at the Economics seminar semin ar at Univ Universit ersity y of Arizo Arizona, na, the Labor and Development Development seminar at Prince Princeton ton University University,, the Public Policy seminar at Berkeley, and the workshop on the “Economics of Crime and Organized Crime” in Palermo, on “Institutions, Individual Behavior and Economic Outcomes” in Alghero, and on the one in Applied Economics in Petr Petralia alia for their useful comments. comments. Marti Martino no Bernardi, Isabella David, Filippo Maggi, and Dominic Smith have provided excellent research assistance. This research was supported by a Collegio Carlo Alberto grant. I wou would ld like to expre express ss my gratitude to the Italia Italian n Academy at Columbia University for their hospitality. † Collegio Car arllo Albe berrto and CeRP, Via Real Collegio 30, Moncalieri, It Itaaly, gi
[email protected]. ‡ © 2011 by Giovanni Giovanni Mastrobuoni. Mastrobuoni. Any opinions opinions expressed here are those of the author and not those of the Collegio Carlo Alberto.
1
Abstract
Using a unique data set on criminal profiles of 800 US Mafia members active in the 1950s and 1960s, and on their connections within the Cosa Nostra network I analyze how the geometry of criminal ties between mobsters depends on family relati rel ationsh onships ips,, com commu munit nity y roots and tie ties, s, leg legal al and illegal illegal act activi ivitie ties. s. I in inter terpret pret somee of the results som results in light light of a model of con connec nectio tions, ns, whe where re more con connec nectio tions ns mean mea n more profits profits but als alsoo a hig higher her risk of defe defecti ction. on. Add Adding ing law enforcem enforcemen entt to th thee mod model el al allo lows ws me to rec recon onst struc ructt th thee sa sampl mpling ing procedur procedure, e, whi whicch I us usee to build buil d a rep repres resen entat tativ ivee sam sample ple of mobs mobster ters. s. I find that variab variables les that lower lower the risk of defection – kinship, violence, and mafia culture among others – increase the number num ber of connec connections. tions. Moreo Moreove ver, r, there is evidence of strate strategic gic endogamy: endogamy: female children are as valuable as male children, and being married to a “connected” woman is a stro strong ng pred predict ictor or of lea leader dershi ship p wit within hin the mafi mafiaa rank ranks. s. A very parsimoni parsimonious ous regression model explains one third of the variability in the criminal ranking of the “men of honor,” suggesting that these variables can be used to detect criminal leaders. One of the rather interesting interesting findings of my simple model is that the distribution distribution of connections is right-skewed, which is remarkably in line with the evidence that mafia organizations tend to be extremely hierarchical. Keywords: Mafia, Networks, Intermarriage, Assortative Matching, Crime. JEL classification codes: A14, C21, D23, D85, K42, Z13
1
Intr In trod oduc ucti tion on
In January 2011, exactly 50 years after Robert F. Kennedy’s first concentrated attack on the American Mafia as the newly appointed attorney general of the United States, nearly 125 people were arrested on federal charges, leading to what federal officials called the “largest mob roundup in F.B.I. history.”1 This paper uses declassified data on 800 Mafia members that were active just before the 1961 crackdown to study the workings of the Mafia, an organization that has been shown to be very hard to eradicate. For example, record number one (shown in Figure 1) is Joe Bonanno. The files tell us that he was born on January 18, 1905 in Castellamare Castellamare (Sicily), was 5 foot 5 inc inches hes tall, and weighted 190 pounds. He became a US citizen in 1945, resided in Tucson (Arizona), and was married to Filippina Labruzzo. He had one daughter and two sons. He had interests in leg legal al bus busine inesse sses: s: Gran Grande de Che Cheese ese Co. Co.,, Fond du Lac (Wisconsi (Wisconsin), n), All Allian iance ce Real Realty ty & Insuran Ins urance ce (T (Tucs ucson, on, Ari Arizona zona), ), and Brun Brunswi swick ck Laun Laundry dry Serv Service ice (Br (Brookly ooklyn, n, New York ork). ). Despite this veil of legal businesses Bonanno was one of the most important Mafia leaders in U.S. He attended all top-level Mafia meetings, including the 1957 Apalachin meeting and the 1956 Bingham Binghamton ton (New York) York) meeting. meeting. He made trips to Ital Italy y to confer with Mafia leaders leaders ther theree and negotiate negotiate for in intern ternati ational onal narcotic narcotic traffi traffick cking. ing. He als alsoo had an extensive criminal history: a record dating from 1930 includes arrests for grand larceny, possession of gun, transportation of machine guns, and obstruction of justice. His closest criminal associates were Lucky Luciano, Francisco Costiglia (Frank Costrello), Giuseppe Profaci, Anthony Corallo, Thomas Lucchese, and Carmine Galante. These records are based on an exact facsimile of the Federal Bureau of Narcotics (FBN) secret files on American Mafia members that were active and alive in 1960 (MAF (MAF,, 2007). 2007 ). I use these files to analyze criminal connections and hierarchies, ranking mobsters based on the number and the quality of their connections. connections.2 I look for empirical evidence 1 2
See The New York Times, January 21, 2011, page A21 of the New York edition. At the end of the 1950s the FBN, which later merged with the Bureau of Drug Abuse Control to form
3
on organizational rules of the Mafia and try to identify the “key players.” I argue that the geometry of Mafia connections is crucial for understanding the activity of the Cosa Nostra (“Our Thing”) as these connections are the building blocks of the entire Mafia, and more general generally ly of organize organized d crime eve even n today today.. Valac alachi’s hi’s 1963 testimony testimony and documents found during the 2007 arrest of Salvatore Lo Piccolo, a Sicilian Mafia boss, show that the first rule in the Mafia decalogu decaloguee stays unchallenged unchallenged:: “No one can presen presentt himself directly to another of our friends. There must be a third person to do it” (Maas (Maas,, 1968). 1968 ).3 Connections are thus necessary for a criminal career within the Mafia. Moreover, leadership positions aren’t simply inherited; soldiers elect their bosses using secret ballots (Falcone and Padovani, Padovani, 1991 1991,, pg. 101). Francisco Costiglia, alias Frank Costello, a Mafia boss connected to 34 mobsters, would say “he is connected” to describe someone’s affiliation to the Mafia (Wolf (Wolf and DiMona, DiMona, 1974). 1974 ). In 1970 the Organiz Organized ed Crime Control Act defined organized crime crime as “The unlawful unlawful activities activi ties of ... a highly organized, organized, disciplined disciplined association....” association....” The purpose of this research research is to shed light on these associations and their organizational organizational structure. structure. The structure has its roots in a world characterized characterized by the absence of legall legally y enforceable contracts. contracts. These criminals need to trust each other, and the purpose of this study is to understand how and where this trust trust eme emerges rges.. Thi Thiss is the firs firstt stu study dy to use network network analysis analysis tools on such a detailed set of information on individual Mafia members to study the emergence of networks: ranging from their business to their family structure. In the 1960s the total estimated number of members was around 5,000 (Maas (Maas,, 1968 1968). ). the Bureau of Narcotics and Dangerous Drugs, was the main authority in the fight against the Mafia (Critchley Critchley,, 2009 2009). ). In New York York the FBI had just four agents, agents, mainly working working in office, assigned to the area, while in the same office more than 400 agents were fighting domestic communists (Maas (Maas,, 1968 1968). ). 3 The remaining 9 rules are: never look at the wives of friends, never be seen with cops, don’t go to pubs and clubs, always being available for Cosa Nostra is a duty - even if one’s wife is going through labor, appointments must be strictly respected, wives must be treated with respect, only truthful answers must be given when asked for information by another member, money cannot be appropriated if it belongs to others or to other families, certain types of people can’t be part of Cosa Nostra (including anyone who has a close relative in the police, anyone with a two-timing relative in the family, anyone who behaves badly and does not posses moral values).
4
Since almost all high-ranking members have a record, the 800 criminal profiles are clearly a non-represen non-repr esentativ tativee sampl samplee of Mafia mem members. bers. A model with endogenous monitoring monitoring highlights ligh ts the ov over-samp er-sampling ling of high-ranking members. members. I devise a wa way y to test the robustness of my findings to random and non random truncation of the network. Several measures on how central and important members are within the network are emplo emp loye yed. d. A mobs mobster ter might might exe exert rt pow power er becau because se he is eit either her connected connected to a larg largee nu nummber of members or connected to a few high caliber figures like Lucky Luciano, Frank Costello, or Joe Bonanno. Mobsters might also be central because they represent bridges that connect different different clusters of a network. network. One of the adv advantag antages es of indice indicess that measure importance based on network-based connections over indices that rely on rank in the organizational hierarchy is that they measure importance in a continuous manner— Mafia bosses are not all equally powerful, and capiregime often differ in their level of importance—, and are more robust to classification errors (Klerks (Klerks,, 2003 2003). ). Guided by a simple model of connections I try to shed light on several hypotheses about how these net netwo works rks and the related related hie hierarc rarchie hiess eme emerge. rge. I ev evalu aluate ate the rel relativ ativee importanc impor tancee of leg legal al and ill illegal egal businesse businesses, s, fam family ily ties, and communit community y ties in sha shapin pingg these networks, contrasting the economic with the social view of Cosa Nostra . Criminals are more likely to be associated with criminals who operate similar illegal businesses busine sses if they try to build cartels. cartels. Or they may try to diversify diversify the risk of detecti detection on by keeping kee ping a low lower er profile and thus associate themselves themselves with criminals criminals who operate differe different nt kinds of businesses. businesses. Careful Carefully ly chosen marriages might might help to establish robust criminal ties. Child Children ren may thus prove prove important, both because of strategic endogamy endogamy and because male descendants descendants represe represent nt trusted poten p otential tial associates associates..4
5
Based on a participant obser-
vation study for a New York based Italian American crime family, Ianni and Reuss-Ianni 4
Information on the number of male and female children, and on the maiden name of the wife is going to allow me to verify the strategic endogamy theory and to evaluate the importance of descendants. 5 Anothe Ano therr ins instru trumen mentt for bui buildi lding ng bond bondss is the “co “compa mparat ratico ico,” ,” a spi spirit ritual ual par paren entage tage a la “Th “Thee Godf Go dfat athe her” r”.. Unfo Un fort rtun unat atel ely y th thee da data ta do no nott co con nta tain in in info form rmat atio ion n ab abou outt th thes esee ki kind nd of li link nkss (Ianni and Reuss Reuss-Ianni -Ianni,, 1972 1972). ).
5
(1972 1972)) discuss the importance of family lineage, intermarriage, and kinship. As in the great Mafia families of Sicily, the “descendants” intermarry continu ti nuou ousl sly y, for the clan defines defines who ma may y or may not ma marry rry whom. whom. As in clans everywhere, the relationship between the intermarrying pair is defined as strengthening the social structure. (Ianni (Ianni and Reuss-Ianni, Reuss-Ianni, 1972 1972,, pg. 192) These marriages would thus have the same function as alliances among European royalty: providing new protectors. Another way children or, more generally, relatives might increase the trust towards a member mem ber is because they represe represent nt potential targets for retaliatory action. action. Trust, or better, blind obedience, and the vow of silence, called omert` a, are indeed essential for a Mafia a clan’s cla n’s surviv survival. Mafia clans, clans, call called ed “F “Fami amilie lies” s” (as in Bonanno (1983 1983)) I use uppercase to dis distin tinguis guish h them from the nu nucle clear ar fami family) ly),, repr represe esent nt socie societie tiess whe where re socia sociall capi capital tal produces public “bads” (Portes (Portes,, 1998 1998). ). A Fami amily ly protects protects its members members and guar guaran antees tees their monopoly power in exchange for part of their revenues. Large clans will therefore be more powerful powerful but also more exposed. A simple model of seque sequentia ntiall “dictator “dictatorial” ial” network network formation by a Mafia boss formalizes this tradeoff and guides my empirical strategy. Some of my economic insights are present in early theoretical analysis of criminal behavio ha vior. r. But most studies studies have have focus focused ed on a mark market et stru structur cturee vie view w of organized organized crime, where the Mafia generates monopoly power power in legal (for a fee) and illegal markets. markets. Among others, such a view is present in the collection of papers in Fiorentini and Peltzman (1997 1997), ), and in Reuter (1983 1983), ), Abadinsky (1990 1990), ), Gambetta (1996 1996), ), and Kumar and Skaperdas (2009 2009). ). Onl Only y tw twoo the theoret oretica icall papers have have focus focused ed on the in intern ternal al orga organiz nizati ation on of organized crime groups. Garoupa (2007 2007)) looks at the optimal size of these organizations, while Baccara and Bar-Isaac (2008 2008)) look at the optimal internal structure (cells versus hierarchies). hierarc hies). I borrow some of their insights, insights, mainly that larger organizations organizations are more profitable but also more vulnerable. 6
Sparrow (1991 1991)) and later Coles (2001 2001)) propose the use of network analysis to study criminal networks. Morselli (2003 2003)) follows their proposal analyzing connections within a single New York based family (the Gambino family), Krebs (2002 2002)) analyzes connections amongg the Sep amon Septem tember ber 2001 hij hijac acke kers’ rs’ terr terroris oristt cel cells, ls, Baker and Faulkner (1993 1993)) study the social organization of three well-known price-fixing conspiracies in the heavy electrical equipment industry, Natarajan (2000 2000,, 2006 2006)) analyzes wiretap conversations among drug dealers, McGloin (2005 2005)) analyzes the connections among gang members in Newark (NJ), Sarnecki (1990 1990,, 2001 2001)) uses network analysis to study co-offending behavior among Swedi Sw edish sh teen teenager agers. s. The These se stu studie diess hig highli hligh ghtt the importance importance of dee deep p tie ties, s, but have have lit little tle or no background information on the individuals. Recent papers on social networks show that an individual’s position within a network is indeed crucial in explaining her or his level of activity (Ballester (Ballester et al., al., 2006 2006). ). If individual decisions are related to the structure of social contacts each individual chooses (or is trapped in), understanding the formation of network structure is crucial for anti-crime policies. Given the complexity of social relationships, evidence documenting patterns of association between agents is a first priority, partly because it can inform the theoretical literature looking at network formation processes.. Sev cesses Several eral models of network network formation have have recently been proposed. Most rely on some form of pairwise regressions (see, for exampl example, e, Bra Bramou moull´ ll´e and Fortin Fortin,, 2010 2010). ). In this study I focus on the elements that are related to the centrality of mobsters rather than determining determi ning single connect connections. ions. I use historical data to understand the functioning of the Mafia, over the last 40 years the Mafia has continued following many of the same rules, and is still active in many countries, including the US. According to the FBI6 , in 2005 there were 651 pending investigations related to the Italian-American Mafia; almost 1,500 mobsters were arrested, and 824 were convicted; of the roughly 1,000 “made” members of Italian organized crime groups estimated estimated to be activ activee in the U.S. U.S.,, 200 were in jail. In addition, addition, the Italian Mafia no 6
The source is www.fbi.gov.
7
longer holds full control of racketeering. With the end of the Cold War and the advent of globalization, “transnational” organized crime organizations are on the rise—mainly the Russian Mafia, the African enterprises, the Chinese tongs, South American drug cartels, the Japanese Yakuza, and the, so called, Balkan Organized Crime groups—and their proceeds, by the most conservative estimates, comprise around 5 percent of the world’s gross domestic product (Schneider (Schneider and Enste, Enste, 2000 2000,, Wagley Wagley,, 2006 2006). ). Williams (2001 2001)) discusses how networks within and across these organizations facilitate their fortunes. Notwithstan Notw ithstanding ding the magnitud magnitudee of these num numbers, bers, the illicit nature of organize organized d crime activities has precluded empirical analysis and the literature has overwhelmingly been anecdotal or theoretical (Reuter (Reuter,, 1994 1994,, Williams Williams,, 2001 2001). ).7 Indeed, this is the first study that uses extensive individual level data to describe the hierarchy of such organizations. I hope that the U.S. experience with Cosa Nostra and this wealth of information may offer clues to promising control techniques in countries where organized crime is on the rise (Jacobs (Jacobs and Gouldin, Gouldin, 1999 1999). ). This work is also related to the growing literature on trust, family values, and family businesses. Guiso et al. (2006 2006)) present an introduction to the importance of culture, defined as “custom “customary ary belief b eliefss and values that ethnic, religi religious, ous, and social groups transmit fairly unchanged from generation to generation,” on economic behavior. I will argue that the same applies to criminal behavior. Bertrand and Schoar (2006 2006)) present a macro-type analysis about the importance of family values for economic growth, and conclude with the comment that more research is needed to understand how family values shape the organization of businesses and their efficiency. I look at how this applies to illegal business. 7
Levitt and Venkatesh (2000 2000)) use detailed financial activities of a drug-selling street gang to analyze gang behavior. behavior. But most gangs do not appear to engage in crimes motivated motivated and organized according according to formal-rational criteria (Decker (Decker et al., al., 1998 1998). ).
8
2
The Th e Ori Origi gin n of Am Amer eric ica an Ma Mafia
Before presenting this empirical study it is important to discuss its historical context. I will talk about when the so-called “made” men came to the U.S. and how the Mafia operated in the 196 1960s 0s wh when en th thee FB FBN N was fil filin ingg the records records I ana analy lyze ze in this study study. Hi Histo stori rians ans define two waves of immigration from Sicily, before and after World War I (WWI). Before WWI immigran immigrants ts we were re mai mainly nly driven driven by economic economic nee needs. ds. Sev Several eral Mafia boss bosses, es, lik likee Lucky Luciano, Tommaso Lucchese, Vito Genovese, Frank Costello, etc, were children of these early immigrants. Even though between 1901 and 1913 almost a quarter of Sicily’s population departed for America, many of the early immigrant families were not from Sicily Sicil y. In that period around 2 million Italians, Italians, mainly from the south emigrated to the U.S. (Critchley (Critchley,, 2009 2009). ). Th Thes esee bab baby y im immi migra grant ntss la late terr bec becam amee st stree reett gan gangg me mem mber berss in the slums; they spoke little Italian, and worked side by side with criminals from other ethnicities, mainly Jewish and Irish (Lupo ( Lupo,, 2009 2009). ). Lured by the criminal successes of the first wave of immigrants, and (quite paradoxically) facilitated by prohibitionism, the second wave of immigrants that went onto to becomee Mafia bosses were already becom already cri crimin minals als by the time they entered entered the U.S U.S.. .. Char Charles les Gambino, Joe Profaci, Joe Bonanno, and others were in their 20s and 30s when they first entered the U.S., and they all came from Sicily. Sicily.8 Another reason for this selection of immigrants was the fascist crack-down of the Mafia, which forced some of these criminals to leave Sicily. After the second wave of immigration the Mafia became more closely linked to the Sicilian Mafia and started adopting its code of honor and tradition.9 In 1930 and 1931 these new arrivals led to a Mafia war, called the Castellamare war, named after a small city city in Sicily where many many of the new Mafia bosses came from. The war lasted until Maranzano, who was trying to become the “Boss of the Bosses,” was killed, 8
Bandiera (2003 2003)) ana analyz lyzes es the ori origin ginss of the Sic Sicili ilian an Mafi Mafia, a, hig highli hligh ghtin tingg ho how w lan land d fra fragme gment ntati ation, on, absence of rule of law, and predatory behavior generated a demand for private protection. 9 See Gosch and Hammer (1975 1975))
9
probably by Lucky Luciano who had joined the Masseria Family. Family.10 This war put Lucky Luciano at the top of the Mafia organization but also led to a reaction by the media and the pros prosecu ecutors tors.. In 1936 Tho Thomas mas E. Dew Dewey ey,, appoi appoint nted ed as New York York City special special prosecutor to crack down on the rackets, managed to obtain Luciano’s conviction with charges on multiple counts of forced prostitution. Luciano served only 10 of the 30 to 50 years yea rs sentenced. sentenced. In 1946 thanks to an alleged invol involvem vement ent in the Allied troops’ landing in Sicily he was deported to Italy, from where he tried to keep organizing “the organization.” Between Betw een 1950 and 1951, the Kefauv Kefauver er Committee, officially the Senate Special Committee Committee to Investigate Crime in Interstate Commerce, had a profound impact on the American public pub lic.. It was the firs firstt com commit mittee tee set up to gain a bette betterr und unders erstand tanding ing of ho how w to fight organized crime, and the main source of information was a list of 800 suspected criminals submitted by FBN’s Commissioner Anslinger, most likely an early version of the records used in this paper (McWilliams (McWilliams,, 1990 1990,, pg. 141). But the Committee could not prove the existence of a Mafia. After Luciano’s expatriation expatriation sev several eral other Famili amilies es headed the organization: Costel Costello, lo, Profaci, Bonanno, Bonanno, and Gambino. Gambino. Family ties ties were of utmost importance. Accordi According ng to Bonanno’ss autobiogr Bonanno’ autobiography aphy (Bonanno (Bonanno,, 1983 1983), ), he became the Boss of the Bosses in part by organizing the marriage between his son Bill and the daughter of Profaci, Rosalia in 1956. In 1957 Gambino took over the leadership. Throughout the 1950s the FBN continued to investigate the Mafia, but until an unexpected event event happened in 1957 the Mafia was not under media attention. The unexpected event eve nt was a lucky raid of an American Mafia summit, the “Apalachin “Apalachin meeting”. Poli Police ce detaine deta ined d ov over er 60 underworld underworld bosses from the raid raid.. Joe Bonanno Bonanno man managed aged to flee flee,, and later came to be known as Joe “Bananas.” After that meeting everyone had to agree with the FBN’s view that there was one big and well organized Mafia. This meant the begin10
Before this event, event, in order to end the power-struggle between Masseria and Maranzano, Lucky Luciano had offered to eliminate Joe “the Boss” Masseria, which he did at an Italian restaurant by poisoning Masseria’s food with lead.
10
ning of the end of the American Mafia. Robert Kennedy, attorney general of the United States, and J. Edgar Hoover, head of the Federal Bureau of Investigations, joined Harry J. Anslinger, the U.S. Commissioner of Narcotics, in his war against the mob. The same years a permanent Senate Select Committee was formed – the McClellan commission. Anslinger’s FBN conducted the investigative work and coordinated nationwide arrests of Apalac Apa lachin hin defendan defendants. ts. Luc Lucky ky Luciano Luciano die died d of a hear heartt atta attack ck at the airport of Nap Naples les in 1962. In the 1950s and 1960s Cosa Nostra was governed by a Commissione of 7-12 bosses, which whic h also acted as the final arbiter on disputes between Famili Families. es. The remaining 10 to 15 families were smaller and not part of Cosa Nostra ’s ’s governing body. After learning that he had been marked for execution Joe Valachi became the first and most important informer for the FBN and later the FBI starting in 1962,11 and revealed that the Cosa Nostra was made of appro approximate ximately ly 25 Famili amilies. es. Each Family Family was was structured in hierarchies with a boss, Capo Famiglia , at the top, a second in command, called underboss, Sottocapo, a counselor, Consigliere , and several capo, Caporegime , captains who head a group of soldiers (regime ) (Maas Maas,, 1968 1968). ). Fi Figu gure re 2 indicates how in 1963, thanks to Valachi’s testimony, a U.S. Senate commission set up to investigate organized crime, called the McClellan commission, drew the Bonanno Family “tree” structure. And Valachi provided information on many more families. Since my data represents a snapshot of what the authorities knew in 1960, they do not contain information about the Family each member belongs to. Nevertheless, we’ll see that the pattern of connections is clearly informativ informa tivee about ab out the structu structure re of Cosa Nostra . Joe Valachi’s testimony confirmed FBN’s view (which at the time wasn’t FBI’s view) that the Mafia had a pyramidal structure with connections leading toward every single member (almost the whole network is connected and the average path length, that is the average distance between two randomly chosen members, is just 3.7). 11
Jacobs and Gouldin (1999 1999)) provide a relati relativel vely y short ov overvie erview w about law enforcemen enforcement’s t’s unpre unpreceden cedented ted attack on Italian organized crime families following Valachi’s hearings.
11
3
The Th e FBN FBN Recor Records ds:: a nonnon-ra rand ndom om sam sampl ple e of mobmobsters
The criminal files come from an exact facsimile of a huge Federal Bureau of Narcotics report of which fifty copies were were circulated within the Bureau starting in the 1950s. These files come from more than 20 years of investigations, and several successful infiltrations by undercover agents (McWilliams (McWilliams,, 1990 1990). ). Given that in the U.S. there were an estimated 5,000 members active during those years the list represents a clearly non-random sample of Cosa Nostra members. More active and more connected mobsters were certainly more likely to be noticed and tracked, which is probably why most, if not all, big bosses that were aliv alive at the ti time me have have a fil file. e. I’ I’m m goi going ng to pr prov ovid idee tw twoo al alter ternat nativ ivee mod model els. s. Th Thee firstt model res firs resem emble bless a sam sampli pling ng proce procedure dure that has been cal called led Res Responde pondent nt-Dri -Driv venSampling (RDS), and is very mechanical. In the second model, instead, law enforcement and mobsters mobsters respond to inc incen entiv tives. es. Both methods methods to rere-we weigh ightt the data lea lead d to very similar conclusions.
3.1
Mark Mar kov-c v-chai hain n Mon Monito itorin ring g
There are no exact records about how the FBN followed mobsters and constructed the network,, though with the use of surv work surveillan eillance ce posts, p osts, undercove undercoverr agent agents, s, etc. agents were were probably discovering discovering previously unknown unknown mobster mobsterss follo following wing known ones. Two photographs taken in 1980 and in 1988 show how these discoveries might have looked like (Figure 3). If one starts with an initial fraction of N of N mobste mobsters rs that are observ observed ed p0 , a 1 × N vector of zeros and ones, called the seed, after k steps the likelihood of observing a mobsters is
p0 T k = pk
(1)
where T is the transition matrix whose columns sum up to one. Given that the seed, the 12
initially observed mobsters, is unknown I use the stationary distribution p, defined as a vector that does not change under application of the transition matrix, or the likelihood that a mobsters has been observed after several steps independently of the seed:
pT = p .
(2)
The Perron-Frobenius theorem ensures that such a vector exists, and that the largest eigen eig env valu aluee ass associat ociated ed wit with h a stoc stochas hastic tic matrix is alw alway ayss 1. For a matr matrix ix with str strictl ictly y positive entries, this vector is unique. The corresponding resampling weights are going to be simply wi0 = p1i , with 0 < pi < 1.
3.2 3. 2
A Model Model of Conn Connec ecti tion onss and Moni Monito tori ring ng
Given the non-randomness of the Cosa Nostra netw network, ork, before b efore describing describing the data and analyzing alyzi ng the determinants determinants of the netw network ork structur structuree I dev develop, elop, based on histori historical cal accounts, a simple simple model of Mafia connectio connections. ns. The aim of the model is twofol twofold: d: i) show show how a simple model of connections can generate a positively skewed distribution of degree and some interesting comparative static results; ii) model the non-random sampling design. In this Sec Section tion I dev develop elop a sim simple ple model of connection connections. s. Eac Each h mob mobster ster decides decides ho how w many connections connections to build. There are no restrictions on the number of connections. connections. Giv Given en that the period betw between een the Castellamare Castellamare war (1929-1931) and 1960 was a relatively relatively peaceful time inside Cosa Nostra , with an established Commissione resolving disputes between Families, I do not model competition or war between Families (Lupo (Lupo,, 2009 2009). ).12 Moreover, after wiping out competing gangs, mainly the Irish and the Jewish, the Italian Mafia had gained gain ed comp complet letee con control trol over over rac racke keteer teering ing.. Duri During ng thos thosee ye years ars the Mafi Mafiaa beha behav ved lik likee a stab stable le cart cartel, el, wit with h sel self-en f-enforc forcing ing strict rules to ke keep ep the cart cartel el runn running ing:: mar marke kets ts we were re 12
See Liddick (1999 1999)) who finds that in the 1960s New York City’s five Mafia Families were jointly running the numbers gambling.
13
geographically segmented, Family memberships were held stable to keep relative power unchanged, the status quo of leadership within Families was endorsed and insurrections were repressed collectively (Bonanno (Bonanno,, 1983 1983). ). For simplicity I also don’t model the choice of different kinds of connections (boss, underboss, captains, and soldiers). 3.2.1 3.2 .1
Law La w en enfor forcem cemen entt
Law enforcement invests in monitoring m at a cost q q and and tries to maximize the probability of disrupting the mobsters connections (1 − p), or alternatively minimize the mobsters’ probability of a successful connection p: N
min
{mi }N i=1
s.t. pi =
pi + qm i ,
i=1
p0i , mi ≥ 0 , q > 0 (1 + mi )γ
p0 i corresponds to mobster’s i initial probability of success, that is without any effort exerted exe rted by the investi investigato gators. rs. The functional functional form for the produ producti ctivit vity y of monitoring monitoring implies that there are decreasing marginal returns in monitoring. The optimal enforcement effort turns out to be
m∗ =
3.2.2 3.2 .2
p0i q
1
γ +1 +1
− 1 if p0i ≥ q
.
(3)
0 otherwise
Optim Opt imal al num number ber of conn connect ection ionss
Mobsters choose their optimal number of connections knowing that additional connections increase their influence but also the risk of betrayal, 1 − p p..13 For simplicity all connections are modeled as if their number was continuous, and they are assumed to increase the mobsters utility or influence by the same amount, normalized to 1. Risks related to each 13
Baccara Bacca ra and Bar-Isa Bar-Isaac ac (2008 2008)) have a game-theoretical model of connections based on information.
14
connections are assumed to be independent. Mobster i’s expected utility is
max spsi for i = 1,...,N , s.t. s > 0 . s
The optimal number of connections is s∗i = −
1 1 =− ln p ln pi ln p ln p0i − ln (1 + m∗ )
1 = − γ ln p ln p0i + γ +1 +1
1 γ +1 +1
ln q
1 < − ln if p0 ≥ q ln p p0i
=−
1 ln p ln p0i
.
otherwise
The number of connections is increasing in the initial probability of a successful and profitablee connecti profitabl connection on p0 . The high higher er the abi abilit lity y of pre preve vent nting ing sol soldie diers, rs, capt captain ains, s, and bosse bossess from becoming becoming informe inf ormers rs for the FBI or the FBN the more connecti connections ons a mobs mobster ter will have. have. What do these abilities depend on? Fear of retaliation for breaking the vow of silence, called omert`a a , has certainly been a key factor for the success of the Cosa Nostra . Som Somee members members were were well known for the violence, their bloodshedding attitude, among these is Albert Anastasia also known as the “Mad Hatter” and “Lord High Executioner,” killed in 1957 in a barber shop because of an internal war between between Vito Genovese Genovese and Frank Costello. Costello. Social ties, in particular, family ties between the boss and his affiliates are also likely to keep p and q high high.. Bos Bosses ses wit with h larg larger er fami familie liess sho should uld therefore therefore ha have ve more con connect nection ions. s. More Moreov over, er, captains with family ties might also have a lower intermediary cost τ . τ .14 In the next Section I present the first empirical test of these hypotheses. Given Giv en the clos closed ed form solution solutionss for s∗ one can easily simulate the distribution of connections for different levels of p of p. Figure 8 shows that even if p if p is drawn from a uniform distribution the model does not predict a uniform distribution of connections, but rather 14
See Reuter (1985 1985)) for a discussion about the optimal size of criminal criminal organizations and, more generally, generally, about providing incentives for loyalty.
15
a positively skewed one. Up until now all mobsters have been treated equally and are supposed to be free to choose an optimal number number of connections. connections. The main reason is that monitoring monitoring needs to be predictable predict able for each mobster in order to contr control ol for the non-random sampling sampling design of the FBN. γ determines the relative importance of this endowment and the cost of monitoring q . If monitoring is very costly the optimal monitoring might be not to monitor at all. 3.2. 3. 2.3 3
Samp Sa mpli ling ng
The presence of law enforcement reduces the variability of the number of connections, but, as long as the returns to investment in monitoring are decreasing, not all the way to zero. Monitoring increases with influence, and more powerful mobsters are more likely to be “sampled.” Solving for the optimal monitoring as a function of the optimal number of connections:
∗
exp −
mi =
= exp −
γ +1 +1 1 γ si
∗
q
1 γ
− ln q
1 1 + ln q γ s∗i
1 1+γ
−1
−1
Given that the minimum number of affiliates is two,15 monitori monitoring ng can be set to 0 whenever whenever s ≤ 2, or
q =
1 = 0.74 . exp1//2 exp1
In order to weight the data based on the optimal monitoring ideally we would need to know γ , but Figure 6 shows that once the weights are normalized to sum up to one, 15
In particular, disregarding the members whose connections have been in part or fully blanked out, Joseph Falcone, Robert Ansoni, Jack Cerone, Frank Napoli, and Mariano Paolacci have just two connections, while none of the mobsters have just one connection.
16
wi =
1
−
mi
N
i=1
1
−
mi
, differences in γ generate very different levels of monitoring, mi , but not
very different weights wi . Not only γ has little effect on the weighting, Figure 7 shows that “Markov chain” weigh we ights ts and “opt “optima imall moni monitori toring” ng” weigh eights ts ha hav ve a sim simila ilarr hy hyperbol perbolic ic sha shape, pe, wit with h the main difference lying where the model predicts no monitoring at all and where we put the monitoring at it minimal level.
4
Desc De scri ript ptiv ive e Evi Evide denc nce e About About Cosa Nost Nostr ra and Its Members
From now on mobsters are going to be described with and without correcting for the non-random non-rand om sampling sampling design. The corrections are done weighting weighting the data based on the weights shown in Figure 6 and 7. Non-weighting corresponds to the case when γ = ∞.
4.1
Indivi Ind ividua duall Chara Characte cteris ristic ticss of the the Members Members
Before analyzing analyzing how crimi criminals nals are connected withi within n the Mafia it is instru instructiv ctivee to describe the members based on the information contained contained in the criminal records. Let me first date the data. Given that the distribution of the year of first arrest has basically full support within the range 1908-1960 (the only year without a first arrest is 1910) one can infer that the data refers to what the authorities knew in 1960.16 The records do not report any deaths, thus thus don’t include big bosses that wer weree killed before 1960, i.e. Albert Anastasia Anastasia boss of one of the 5 New York City families, the Gambino family. family.17 Table 1 shows that the average age is 52 years, with the oldest mobster being 81 years old, and the youngest one 23. Corr Correcti ecting ng for the sam sampli pling ng wit with h tw twoo very different different γ s–1 s–1 and 15–decreases the 16
Additional evidence Additional evidence is the follo following wing description description in Mic Michael hael Russo’s file: “Rece “Recently ntly (1960) perjured himself before a Grand Jury in an attempt to protect another Mafia member and narcotic trafficker.” 17 His brother Anthony “Tough Tony,” instead, was killed in 1963 and is in the records.
17
averag av eragee age by very little. little. Hal Halff of the mobsters mobsters reside reside in either either New York York,, or in New Jersey, and probably entered the U.S. through Ellis Island. Indeed, 29 percent were born in Sicily Sicily and another another 10 perce percent nt in othe otherr regi regions ons of Ital Italy y. Prope Properly rly weigh weightin tingg the data fewer turn out to be born in Sicily, as Sicilians tend to have leadership positions and are thus oversampled, irrespectively of the weighting procedure that is used. Most remaining mobsters were born in the United States but were of Italian origin as this was a prer prerequ equisi isite te to become a mem member. ber. 73 percent percent of members members are mar married ried (70 percent when weighting), but only 60 percent of these are reported to have children. The overall average number of children is 1 and is 2.14 among members with children, equally divide div ided d betw between een sons and daug daught hters ers.. 18 perce p ercent nt of mem members bers are marr married ied to som someon eonee who shares her maiden name with some other member (Connected wife ), ), though fewer are when weighting (15 percent when the economic model is used, 14 percent when the Markov Mark ov chain weights weights are used). These marriages are presumably endogamous within within the Mafia.. Obs Mafia Observ ervee that I’m understati understating ng the percentage percentage of marr marriag iages es within the Mafi Mafiaa as some Mafia surnames might be missing in the data data..18 The FBN reports an average of 1.97 siblings per member, while the average number of recorded members that share the same surname is 1.62. These numbers tend to be smaller when weighting. The average height is 5.6 feet, the average weight is 176 pounds, no matter how the weighting is done.19 Mobsters’ criminal career starts early. They are on average 25 years old when they end up in jail for the first time, and the majority has committed some viol vi olen entt cri crime me.. On Only ly 16 perc percen entt do not ha hav ve an arrest arrest record record.. I do don’ n’tt kn know ow the total number of crimes committed by the mobsters but I know in how many different types of crime they have apparently apparently been inv involv olved. ed. This number number varies between between 0 and 9 and the average is 2.58. I also know the number of different legal businesses they have interest in. This number varies between 0 and 5 and is on average equal to 1. 18
While it is also possible that some women might have a Mafia surname without being linked to any Mafia family, this is very unlikely conditional on being married to a Mafia associate. 19 As a note, 18 percent of the mobsters are obese and 58 percent overweight.
18
Another variable summarized in Table 5 is the interaction index . The index measures the exposure to what Bonanno (1983 1983)) calls, in uppercase, “Tradition” or Hess (1973 1973)) calls, in lowercase, “mafia,” or “mafia culture” to distinguish it from “Mafia” the organization. Looking at Figure 9 hel helps ps explaini explaining ng the ind index. ex. It shows shows the current current dis distri tributi bution on at the zip code level of the members’ surnames in Italy’s phone directory. directory.20 Each circle is proportion propor tional al to the number number of sur surname namess pres presen entt wit within hin each zip code. Not surprisin surprisingly gly many man y sur surnam names es sho show w up in Sic Sicily ily,, in Naples, Naples, and in Cal Calabri abria. a. Man Many y of these surnames surnames appear also in large cities that were subject to immigratory flows from the south, like Milan, Rome, and Turin. For each members’ surname I computed the probability that it shares a randomly chosen zip code located in the South of Italy with other surnames from the list. list. To be more pre precis cise, e, the ind index ex for mem member ber i is equal to 100,000 times the sum across zip codes j of the fraction of surnames of member i present in zip codes j times the fraction of surnames of the other members (−i) in the same zip code:
interactioni = 100, 100, 000
j
#surnamei,j j #surname i,j
#surname−i,j . # surname i,j − j
(4)
The advantage of this index is that I can computed it for all surnames while information about the Italian community of origin would only be available for those born in Italy. The average index is equal to 2.67 per 100,000, though it’s lower when weighting the data (2.01). (2.0 1). Ten percent percent of the times the index is zero, either either because the zip codes do not overlap or because the surname is not in the phone directory. Tables 2 and 3 show the list of legal and illegal activities that at least 5 percent of members mem bers we were re in invo volv lved ed in. Weigh eightin tingg does lit little tle to the distributi distribution on of leg legal al and illegal illegal activities. Most mobsters owned restaurants, drugstores or were otherwise involved with the supply supply of food. Real estate, estate, cas casinos inos,, car dealershi dealerships, ps, and impor import-ex t-export port were also common com mon business businesses. es. Drug traffickin traffickingg is the mos mostt com common mon of the illegal illegal activitie activitiess (43 20
Unfortunately I could not find the distribution of surnames in 1960.
19
percent perce nt), ), perhap perhapss part partly ly becau because se the inf inform ormatio ation n wa wass gath gathered ered by the FBN FBN.. Tw Twen enty ty-six percent percent of mem members bers we were re in invo volv lved ed in robberies robberies and 23 perce percent nt in mu murde rders. rs. Weapon offences offen ces and ass assaul aults ts are also qui quite te comm common. on. Some crimes crimes that are ty typic picall ally y ass associat ociated ed with organized crime, like gambling, extortions, and liquor offences (during prohibition) are highly represented as well.
5
Validi alidity ty of Net Netwo work-ba rk-based sed Meas Measures ures of Importa Importance nce
Before moving towards the second part of the paper, where I try to predict mobsters’ power inside the criminal organization, I’m going to validate the use of network-based power pow er meas measures ures.. In the introducti introduction on I men mention tioned ed that connectio connections ns we were re the building building block of the Mafia, and that only directly connected mobsters were allowed to openly talk about their secret organization. Each criminal record contains a list of criminal associates. Figure 1 indicates, for example, that Joe Bonanno was associated with Luciano, Costello, Profaci, Corallo, Lucchese, and Galante. There is no evidence about how the FBN established such associations, and why they were restricted to be 5. But it seems apparent that each profile lists its highest associates.. Notice that the mere keeping track of connections associates connections shows that even the FBN understood how important these connections were. Indirected connections are clearly more numerous, as mobsters can be listed as associates in several records. Hence, I define two mobsters to be connected whenever at least one mobster lists the other mobster’s last name in his record.21 Table 4 shows the list of members with the 10 highest and 10 lowest number of direct connections.22 For each mobster the FBN records contain a paragraph about his activities within Cosa Nostra . In 21
In other words, I construct a symmetric adjacency matrix of indirected connections. Since Sin ce con connec nection tion are bas based ed on sur surnam names es mem members bers wit with h the sam samee sur surnam namee wil willl sha share re the sam samee This is introd introduc uces es some noise noise but de deal alin ingg wi with th th thee la larg rgee var aria iati tion on in fir first st names, names, i. i.e. e. An Antotodegree . Th nio/Tony/Anthony, would introduce even more noise in graphing the network. 22
20
order to extract information on the level of importance of these criminals the variable Top counts the number of times the words “boss,” “highest,” “most,” “head,” and “top” are cited and High the number of times “high,” “influential,” “important,” “leader,” “leading,” “powerful,” and “representing” are cited. The Apalachin variable indicates whether the mobster attended the important 1957 Mafia meeting in Upstate New York. The last column represents the historically reconstructed position within the Mafia. Criminals with many connections are more likely to be recognized as high-ranked members, mem bers, and more likely likely to hav havee attended the 1957 meeting. Sev Several eral members in the top distribution of the number of connections are bosses, i.e. Salvatore Lucania, alias Lucky Luciano, Vito Genovese, Antoni Accardo, Joe and Joseph Profaci. Salvatore Santoro and Salvatore Vitale were instead underbosses of the Lucchese and Bonanno Family. Criminals with the lowest degrees , instead, are mostly soldiers. The number of connections, called degree in network analysis, is clearly the easiest way to measure the importance of members, but in recent years social network theorists proposed different centrality measures to account for the variability in network location across acro ss agen agents, ts, and there is no sys system tematic atic criterium criterium to pic pick k up the “rig “right ht”” cen central tralit ity y measure for each particular situation (Borgatti (Borgatti,, 2003 2003,, Wasserman and Faust, Faust, 1994 1994)).23 Before showing the distribution of some of these measures I briefly describe them (see Appendix A for a more technical definition). Unlike degree , which weights every contact equally, the eigenvector index weights contacts according to their centralities. The index takes direct as well as indirect connections and thus the whole network into account.24 The closeness index represents the inverse of the average distance between a node (a member) and all the other nodes, and is a good measure for how isolated members are. The betweenness index measures the number of times a node is on the shortest path between two randomly chosen nodes, and is a good measure for the member’s capacity to 23
See also Sparrow (1991 1991)) for a discussion on centrality indices in criminal networks. As first noted by Granovetter (1973 1973), ), we weak ak ties (i.e. (i.e. fri friend endss of friends) friends) are important important source source of information. See Patacchini and Zenou (2008 2008)) for the role of weak ties in explaining criminal activities. 24
21
act like a bridge between clusters of members. These measures of individual centrality allow me to consider different nuances in the definition of a Mafia leader. The data also allows me to construct a qualitative indicator of importance, based on the union of the Top and High variables discussed in this section. Table 5 collec collects ts descri descriptiv ptivee statis statistics tics about ab out various indicators of importance. importance.25 Weighting the average number of connections mobsters have drops from 11 to 7, and more generally all centrality measures decrease once we control for the non-random nature of the sampling design, no matter the weighting procedures that is used. The non-standardized degree varies between 1 and 71, and Figure 5 demonstrates that the correspondi corresponding ng density is positiv positively ely skewed. skewed. The eigenvector eigenvector index (cen (centralit trality) y) has a density that is very similar to that of degree , while the density of closeness is more symmetrically distributed, meaning that most mobsters are neither too isolated nor too close within within the netwo network. rk. The densit density y of betweenness , instead, shows that very few mobsterss represe mobster represent nt bridges between subsets of the netw network, ork, most likely Families Families.. Finall Finally y, fifty-five percent of the mobsters appear to be high-ranked, meaning that their descriptions includes at least a word that implies leadership. The list of the names of mobsters with the highest and the lowest degree provides only limited limited evidence that connections matter. For this reason I used historical historical sources to collect information on the highest ranks achieved by the mobsters up until these days. Ideally one would like to reconstruct their rank in 1960, but this tuned out to be impracticable. practica ble. Table 6 shows that 70 percent of mobsters are soldiers or their rank could not be foun found. d. Both type of mobs mobsters ters have have the lowest lowest centrali centrality ty measures measures,, and degree is close to 10, which happened to be the typical number of soldiers under the control of captains, called “capodecina” literally “head of ten.” The third largest group of mobsters are bosses, which suggest that over time several underbosses, captains, and Consiglieri becamee “don becam “dons.” s.” Bos Bosses ses and und underbos erbosses ses ha hav ve the hig highes hestt betw between eennes nesss ind index, ex, whi which ch is 25
The individual centrality measures have been normalized to be between 1 and 100
22
consistent with bosses acting as bridges across Mafia Families. Consiglieri , instead, have the highest degree, which is consistent with the role of counselor for the entire Family, but with little bridging capacity across Families. Different measures of centrality thus seem to capture different roles within the criminal organization. Information on the place of residence of the mobsters allows me to perform a final validity test of centrality measures. One would expect more powerful mobsters to livee in richer liv richer hom homes. es. Whi While le ideally ideally one would would lik likee to have have inf informa ormatio tion n on the valu aluee of these homes in 1960, I could only collect information information on their value value today. today. Figure 4 shows that the value is increasing with network centrality, and Table 7 shows that this increase is robust robust to the inc inclus lusion ion of sta state te fixe fixed d effe effects cts and othe otherr regr regress essors. ors. The Table Table shows shows the coefficients from a regression of the log eigenvector index on the log of the house valu alue. e. Esti Estimate mated d elas elastic ticiti ities es are clos closee to 10 perce percent nt.. The last three columns columns replicate replicate the regressions using as a measure of importance a dummy equal to one when mobsters descri des cripti ption on con contai tains ns one of these wo words rds:: boss boss,, hig highes hest, t, mos most, t, head head,, or top. Cond Conditi itional onal on state fixed effects the coefficients are less precisely estimated, suggesting that network based measures might perform better measures based on the description of the mobsters.
6
Whic Wh ich h Chara Charact cter eris isti tics cs Predi Predict ct the Importa Importanc nce e of a Mafia Member?
Now that connections have been shown to matter, the next step will be to highlight the forces that might influence p, and, therefore, their number. While for the sake of simplicity the model does not take into account the quality of the links, that there are several ways to measure measure ho how w cen central tral,, or how conn connecte ected d a mem member ber is within within the network network.. I sta start rt my regression analysis using the eigenvector index as dependent variable, as it depends both on the number and the quality of its connections, on the whole sample of mobsters.
23
6.1 6. 1
Desc De scri ript ptiv ive e Ev Evid iden ence ce
Before moving to the regressions a graphical analysis can shed light on some organizational rules rul es of the Mafia. Fig Figure ure 10 shows that all centrality measures grow steadily with age untill age 60 and later starts decreasing. unti decreasing. Lik Likee typical earning profiles there is an inv inverse erse U-shaped relationship relationship between centralit centrality y and age, but while for earnings profiles the peak p eak is typical typically ly around 45 or 50, in the Mafia . Thi Thiss find finding ing confirms confirms Ianni and Reuss-Ianni (1972 1972,, pg.130)’s anthropologic results about the importance of age in determining the leaders lea dership hip posi p osition tions. s. The diff differen erence ce betw between een the minimum minimum age (23) and the max maxim imum um age (60) doubles the degree, triplicates the eigenvector index, and more than triplicates Closeness, ness, instead, shows shows a very steep increase up to age 30 and then flattens betweenness . Close out, but the overall increase is more modest. Figure 11 presents how the interaction index, the continuous proxy for “mafia culture” influences centrality. The overall patterns are less clear than for age, but when interactions are very high all centrality centrality measures appear to increase as wel well. l. Eigen Eigenvec vector tor centrality centrality is low in the absence of daughters or sons, but sons and daughters seem to be equally valuable in increasing the father’s centrality (Figure 12 12). ). I will say more about this later when whe n I anal analyze yze the impo importan rtance ce of in interm termarri arriage. age. Fig Figure ure 13 shows that the number of types of crimes and businesses also positively influence mobster’s centrality within the netwo net work, rk, especially especially up to the 4th type. Abo Above ve 4 ty types pes of crimes crimes all alleged egedly ly committed committed the evidence is more noisy as fewer observations observations are available. available. I present how categorical variables influence network centrality in the next Section, where I move to a regression framework.
6.2 6. 2
The Ei Eige gen nval alue ue In Inde dex x
In Table 9 I use the eigenvector centrality measure as the dependent variable, though more generally I assume the following linear relationship between the k-th index of importance 24
cki of indiv individual idual i and his observable characteristics X i : cki = β ′X i + ei , k = 1, .. ..., ., 4.26 Each regression controls for the number of mobsters that in the data share the same surname sur name.. I cal calll thi thiss varia ariable ble “Extended “Extended fam family ily members” members” despite despite the poss p ossibi ibilit lity y that somee of these mobsters som mobsters mig might ht not be rel related ated to eac each h othe other. r. Giv Given en that the connecti connections ons are based on surnames I add that this variable is to control for the mechanical effect that an increase in the number number of mobsters might have have on the centrality centrality measures. measures. The coefficient is around, which is also the average number of associates contained in each criminal crimin al file. In the first column I regress the eigenvector index on individual characteristics that, desp de spit itee the possibi possibili lity ty of se sele lecti ction on bi bias as,, are less li lik kel ely y to be en endog dogeno enous us.. Ag Agee en ente ters rs quadrat qua dratica ically lly.. Pe Peak ak cen central tralit ity y is reached reached at age 59 and both term termss are significan significant. t. The finding of a peak late in life is consistent with Ianni and Reuss-Ianni (1972 1972)’s )’s account of hierarchies based on generations (though it might also reflect that more able mobsters are more likely to be alive). Weighting does little to the age at which there is a peak. Another variable that has a significant positive influence on centrality is being born in Sicily (a 20 percent increase), independent of the weighting used. This means that not only is the American Mafia an Italian enterprise, it values direct links with Sicilians more than with people from other parts of Italy. Italy. Sicil Sicilian ian kin-centered kin-centered social system, with its code of honor and vow vow of silence, silence, form formss the buildin buildingg bloc block k for the Mafia Mafia.. Bon Bonanno anno would would write that among Italians Italians he felt safe only around Sicilians. Sicilians. Moreo Moreover ver,, nativi nativity ty does not fully capture capture adherence to the mafia code of law as 60 percent of the mobsters were born in the U.S. This is why I use the interaction index, which depends on the geographical distribution of last names, to proxy for community ties and exposure to the “Tradition.” The index has a positive and significant effect on network centrality, especially after weighting the data.. Mem data Members bers from Sic Sicily ily and those with str stronge ongerr tie tiess to the Mafi Mafiaa cul culture ture are lik likely ely a, increasing p. Body a to be trusted more, as they are more likely to adhere to the omert` 26
See Appendix A for a detailed description of the different indices.
25
weight of the mobster does not influence his status, while his height does but only after weighting the data. In column 4, 5 and 6 I control for additional factors, like the nuclear family structure of the mobster, that might in part be endogen endogenous. ous. The purpose is thus merely descriptiv descriptive, e, and useful for predic predictiv tivee purposes. Each additional additional child increases the eigenvector eigenvector index by 0.8, but the effect gets weaker weaker when weighti weighting ng the data. Child Children ren thus seem to be more valuable for higher ranked mobsters. It doesn’t matter whether the child is male or female. This finding challenges the criminological view that within the Mafia male children are more valuable than female ones because they represents potential “workforce.” The leading explanation for this finding is that in a male only society like the Mafia “connected” girls (probably in excess demand) could be married strategically. The number of siblings, and being married or divorced does not influence the centrality index. ind ex. Thi Thiss is in line with with Falcone and Padovani (1991 1991,, pg. 113) 113)’s ’s view view that unlike unlike the Mafia in Ital Italy y the American American Mafia adop adopted ted a mor moree libe liberal ral view towards towards divorce divorce.. Bei Being ng married to a wife who is connected, instead, leads to an almost 50 percent increase in the index (slightly (slightly less when weighting). weighting). These findings, findings, again, seem to suggest that higher p,, due to more trusted p trusted links, links, inc increas reasee the optimal optimal number number of connectio connections. ns. I cann cannot ot rul rulee out the possibility that more connected mobsters are also more likely to find, or be given, a conn connecte ected d wif wife. e. Whi While le these alt alterna ernativ tivee in interp terpreta retation tionss do not matter when the only purpose is to discover the leading figures within the Mafia, one has to be careful in giving a causal interpretation interpretation to these estimates. estimates. The leading mobsters are active in New York and New Jersey. The remaining two variables measure how many different types of crime and different different types of businesses the mobster was involv involved ed in. Both variables variables are positiv p ositive. e. Thiss is consisten Thi consistentt wit with h more able criminal criminalss being more able to div divers ersify ify risk. But the number of business types (+16 percent) are a better predictor for “key players” than the number of crime types (+4 percent). The combined regressors explain almost 30 percent
26
of the variability of the eigenvector index. Criminals who were known to have committed violent crimes have a larger index, but the effect is not significan significant. t. If being vio violen lentt redu reduces ces the ris risk k of betrayal betrayal this findi finding ng is consist cons isten entt wit with h my model. An arrest inc increas reases es the importance, importance, thou though gh agai again, n, the effect is not significant, and endogeneity might mean that higher ranked individuals might be more mo re li lik kel ely y to be arr arres este ted. d. Ag Agee at firs firstt arr arres estt is negativ negative, e, sh shoowi wing ng tha thatt not only age, but also experience increases the level of centrality in the network (young members are often recruited in jail; Ianni and Reuss-Ianni, Reuss-Ianni, 1972 1972,, pg. 45) 45),, but this effec effectt is also also not significant once I control for all the other factors.
6.3
Other Oth er Net Netw work Cen Centra tralit lity y Indices Indices
Eigenvector centrality represents only one way to measure centrality in the network. Other indices capture different nuances of centrality. In Table 10 I look at different measures of centrality and importance using the same specification used in the last column of Table 9. For brevity I only show the results that control for monitoring with γ = 15. In order to highlight how different measures are able to capture different characteristics of the network, and since there is no systematic way to decompose the different centrality meas me asure uress bas based ed on the im import portanc ancee of di direc rectt and in indi dire rect ct li link nks, s, I de dev vel elope oped d a si simp mple le statist stat istica icall wa way y to acco accompl mplish ish the sam samee goal goal:: I sim simply ply add as a depen dependen dentt varia ariable ble the residual (ǫ (ǫ ji ) of a linear projection of the alternative measures of centrality (c (c ji ; j = 1) on j
j
= 1. The coefficients on the residuals of a particular degree (c1i ): ci = α + βc 1i + ǫi ; j centrality index measure the nuances captured by that particular index with respect to a simple count of direct connections. connections. Colum Column n 2 replic replicates ates the last column of Table Table 9 to easee the comparis eas comparison on wit with h the other measures. measures. Com Compari paring ng columns columns 1 and 2 sho shows ws that the coefficients are not very different when degree is used to proxy for leaders leadership hip.. The only coefficients that seem to be smaller are the ones on residing in New York or New
27
Jersey, and the one on the types of businesses. A simple test for the significance of these differences that looks at the residuals confirms that these differences are significant. This means that living in New York and New Jersey, and the number of businesses increase the eigenvector centrality not only through the direct links but also through the indirect links, while for all the other variables that are significant in column 2 but not in column 3 only direct links matter (connected wife, age, and the interaction index). Closeness , an inverse measure of the average distance from the other members, and betweenness , the ability to build bridges, capture different aspects in the definition of
indi in divi vidua duall im import portanc ancee or pow power er wi withi thin n a net netw work ork.. Co Colu lumn mn 4 sh show owss th that at ov over erall all th thee coefficients for closeness do not differ substantially from those seen before. Closeness is more than any other index dependent on kinship. The “Sicily” coefficient, the interaction index coefficient, and the “Connected” wife coefficient are all highly significant. Moreover, column 5 shows that a large part of these effects is driven by the indirect links (cannot be explained by degree alone). Results that use betweenness as a dependent variable shown in column 6 present a very ver y interesting interesting findings: there is no eviden evidence ce that bridges are build through marriages. The coefficient on the Connected wife dummy is precisely estimated to be close to zero. Strategic marriages are thus confined to happen within “friendly” clans, and not across clans that wouldn’t wouldn’t otherwise otherwise be b e connected. The number number of different different crimes committed also have no bridging capacity, while businesses do. Experiencee and age, instea Experienc instead, d, influen influence ce betweenness beyond the increase in degree . And so does being Sicilian, residing in New York or New Jersey, and being active in several businesses.
28
6.4
Qualit Qua litati ativ ve Measu Measures res of Import Importanc ance e
In column 2 of Table 11 I exp exploi loitt qua qualit litativ ativee inf informa ormatio tion n on the mob mobster sterss con contain tained ed in the records. records. In particula particular, r, I use the same specificati specification on as befor beforee but the dependent dependent variable counts the number of times the following words are used in describing a mobster: “boss,” “highest,” “most,” “head,” ‘top,” “high,” “influential,” “important,” “leader,” “leading,” “powerful,” and “representing.” Overall the results confirm the importance of directly migrating from Sicily, being old and experienced, residing in New Jersey or New York York,, being active active in man many y ty types pes of bus busine inesse sses, s, and using violence violence.. But marrying marrying a connected wife, having children, having strong community ties (interactions), and the number num ber of crimes committed committed are no longer significan significantly tly related to leadership. This lack of significan significance ce lea leads ds to a lo low wer R-sq R-square uared d (8 versus versus 28 perce percent nt). ). Whi While le there might might be superior ways to extract the information know by the FBN (we did try several qualitative measures), network based measures seem to capture additional attributes of leadership.
6.5
Indegree and Random Truncation of the Sample
Table 12 shows two types of robustness checks. In the first instead of using indirected links I use only the indegree , meaning the number of times someone appears as an associate in crimin cri minal al profiles profiles of othe otherr mob mobste sters. rs. Giv Given en that outdegree is somehow bounded it is not surprising that the coefficients are driven by indegree , and thus are very close to the ones based on indirected connections shown in the first column of Table 10. The second and more important robustness check simulates random truncations of the sample. Since the initial sample is likely to be truncated as well—some scholars estimate that there are as many as 5,000 “made” men across the U.S—, I try to address the possible bias that such a truncation might generate (see Borgatti et al., al., 2006 2006,, for a similar simulation). Columns 3 and 4 show the mean and the median coefficients obtained in regressions based on 500 different samples where I randomly truncated 50 percent of the observations. Indegree is 29
based on 50 percent of the sample only and, as before, standardized to lie between 0 and 100. Truncatio runcation n tends to reduce the size of the coefficien coefficientt on the dummy “connected “connected wife,” which is reasonable given that smaller samples lead to an increase in misclassification error of that particular variable. The other coefficients tend to be quite stable.
7
Conc Co nclu ludi ding ng re rem mar arks ks
This paper presents the first thorough micro-level analysis of the US Mafia network. Beside testing sociological and historical views about the functioning of these criminal networks, I develop a simple model of connections that highlights the trade-off of connections: tion s: they increas increasee profi profits ts and pow power er but als alsoo the risk of dete detectio ction. n. More connecti connections ons mean mea n more potential potential informer informers. s. Moni Monitori toring ng by law enforceme enforcement nt is shown shown to depen depend d on the number of connections, which allows me to control for the non-random sampling in the empirical section. Family ties, violence, and exposure to Mafia culture reduce the probability of defection and increas increasee the number of connecti connections. ons. Unlik Unlikee economi economicc organiza organizations tions hierarchies hierarchies depend crucially on kinship. My results highlight how intermarriage shapes the network. Women are used to foster the Family’s network centrality, but only within trusted Families. ili es. Woman are not used to bri bridge dge Fami amilie liess that are not otherwise otherwise closely closely conn connecte ected. d. Trust shapes the network. Where values are shared and the mafia culture is strong, connection nec tionss are more stab stable le and th thus us more nu numer merous. ous. Cohe Coheren rentt wit with h mafia culture, culture, ho how w members bers are within within the network network inc increas reases es ste steadi adily ly with age. And more cen central tral central mem members have more businesses, legal and illegal ones. If social connections are the driving forces of the phenomenon under consideration and if their structure is non-random, as in my case, a detailed study of the characteristics of the network netw ork reveals some relev relevant ant features of social structure that can guide crime prevention prevention policies. A targeted policy identifying “key players” in a given area may be an effective way 30
to reduce crime (Ballester (Ballester et al., al., 2006 2006,, R´eka eka et al al.., 2000 2000). ). A key player player is an ind indivi ividual dual belonging to a network of criminals who, once removed, leads to the highest aggregate delinq del inquen uency cy redu reducti ction. on. In prac practice tice,, the plan planner ner ma may y wa want nt to ide ident ntify ify optimal optimal net netwo work rk targets targ ets to conc concen entrat tratee (sc (scarce arce)) in inv vest estigat igatory ory res resourc ources es on som somee part particu icular lar ind indivi ividual duals, s, or to isolate them from the rest of the group, either through leniency programs, social assistance assis tance programs, or incarceration. incarceration. A necessary condition condition for design designing ing such policies is the ability to map and identify a social network structure. Detailed information about the hierarchy might even allow enforcement agents to break the chain of command by arresting sets of mobsters (Farley (Farley,, 2003 2003). ). Only detailed detailed empirical studies studies on real-w real-world orld social networks can provide guidance in this direction.
31
A
Cen Ce ntr tral alit ity y Me Meas asur ures es
Let N = {1, . . . , n} be a finite set of agents in network g. Le Lett me defi define ne G as the n−square adjacency matrix of the network g , i. i.e. e. the matri matrix x tha thatt kee eeps ps track track of th thee direct dir ect conn connecti ections ons in thi thiss net netwo work, rk, whe where re gij = 1 if i and j are dir directl ectly y lin linke ked, d, and gij = 0, otherwise. otherwise.27 The simplest index of connectivity is the number of direct links stemming from each agent i in the network, i.e. degree centrality: n
1 i
c =
gij .
=1 j=1 j
The definition of centrality is thus based on the number of direct links only. A variant of simple degree is eigenvector centrality, which also takes into consideration indirect links: n
c2i = λ−1
gij c j2
j=1 j =1
where λ is the highest eigenvalue of matrix G. The formula formula implies implies (rec (recursi ursive vely) ly) that the centrality of individual i is proportional to the sum of centralities of the individuals she/he is connected to. It thus can be high even if she/he has low degree . The standard measure of closeness centrality of individual i is given by: c3i =
1
n j=1 j =1
dij
where dij is the geodesic distance (length of the shortest path) between individuals i and j..28 As a result, the closeness centrality of individual i is the inverse j inverse of the sum of geodesic 27
Measurement errors in the definition of the actual connections are clearly unavoidable in this study. Some names are blanked out and many lower-ranked mafiosi were not filed. 28 The length of a shortest path is the smallest k such that there is at least one path of length k from identify y suc such h a length by computin computingg G, G2 , G3 , ..., until I find the first k such that the i to j . I can identif (i, j )th entry of Gk is not zero.
32
distances from i to the n − 1 other individuals and can be regarded as a measure of how long it will take information to spread from a given member to other members in the network. Betweenness indexes derive from the number of optimal paths across (or from) every node. It can be defined as: n
4 i
c =
j,l
a jl,i a jl
where j and l denote two given agents in g, a jl,i is the number of shortest paths between j and l through i, and a jl is the number of shortest paths between j and l.
33
References MAFIA: The Government’s Secret File on Organized Crime. By the United States Treasury Department, Bureau of Narcotics . HarperCollins Publishers, 2007.
Howard Abadinsky. Organized Crime . Wadsworth Publishing, 1990. Mariagiovanni Baccara and Heski Bar-Isaac. How to Organize Crime. Review of Economic Studies , 75(4):1039–1067, 2008.
Wayne E. Baker and Robert R. Faulkner. The Social Organization of Conspiracy: Illegal Networks in the Heavy Electrical Equipment Industry. American Sociological Review , 58(6):pp. 837–860, 1993. URL http://www.jstor.org/stable/2095954 . Coralio Ballester, Antoni Calv´o-Armen o-Armengol, gol, and Yv Yves es Zenou. Who’ Who’ss who in net netwo works rks.. wanted: The key player. Econometrica , 74(5):1403–1417, 09 2006. Oriana Bandiera. Land reform, the market for protection, and the origins of the Sicilian mafia: theory and evidence. Journal of Law, Economics, and Organization , 19(1):218, 2003. Marianne Bertrand and Antoinette Schoar. The role of family in family firms. Journal of Economic Perspectives , 20(2):73–96, Spring 2006.
Joseph Bonanno. A Man of Honor: The Autobiography of Joseph Bonanno . St. Martin’s, 1983. Stephen Step hen P. Borgatti, Borgatti, Kath Kathlee leen n M. Carl Carley ey,, and Da David vid Krackha Krackhardt. rdt. On the Robu Robustn stness ess of Centrality Measures Under Conditions of Imperfect Data. Social Networks , 28(2): 124–136, 2006. Steve P. Borgatti. The Key Player Problem , pages 241–252. R. Breiger, K. Carley and P. Pattison, Committee on Human Factors, National Research Council, 2003. 34
Yann Bramo Bramoull´ ull´e and Berna Bernard rd Fortin ortin.. The Econometrics of Social Networks . New Palgrave Dictionary of Economics, steven n. durlauf; lawrence e. blume edition, 2010. Nigel Nig el Coles. Coles. It’ It’ss Not What You Kno Know-I w-It’s t’s Who You Kno Know w Tha Thatt Cou Count nts. s. Ana Analys lysing ing Serious Crime Groups as Social Networks. British Journal of Criminology , 41(4):580, 2001. David Critchley. The Origin of Or Organiz ganizeed Crime in Amer Americ ica: a: The New York City Mafia Mafia,, 1891–1931. Routledge, 2009.
Scott Sc ott H. De Deck cker, er, Ti Tim m By Byn num um,, and Deborah Deborah Wei Weise sel. l. A Tal alee of tw twoo Ci Citie ties: s: Gan Gangs gs as Organized Crime Groups. Justice Quarterly , 15(3):395–425, 1998. Giovanni Falcone and Marcelle Padovani. Cose di Cosa Nostra . Rizzoli, 1991. Jonathan D. Farley. Breaking Al Qaeda cells: A mathematical analysis of counterterrorism operations (A guide for risk assessment and decision making). Studies in Conflict & Terrorism , 26(6):399–411, 2003.
Gianluca Gianlu ca Fioren Fiorentini tini and Sam Pel Peltzman. tzman. The Economics of Organised Crime . Cambridge University Press, 1997. Diego Gambetta. The Sicilian Mafia: the business of priva Harv vard Univ Univ private te protec protection tion . Har Press, 1996. Nuno Nu no Ga Garo roup upa. a.
Opti Op tima mall la law w en enfo forc rcem emen entt an and d cr crim imin inal al or orga gani niza zati tion on..
nall of Ec na Econ onom omic ic Be Beha havi vior or & Or Orga ganiz nizat atio ion n , 63 63(3 (3): ):46 4611–4 –474 74,, July 200 0077.
Jour-
URL
http://ideas.repec.o http://ide as.repec.org/a/eee/jebor rg/a/eee/jeborg/v63y2007i3p461 g/v63y2007i3p461-474.html -474.html .
Martin A. Gosch and Richard Hammer. The Last Testament of Lucky Luciano . Litt Little, le, Brown, 1975.
35
Mark S. Granove Granovetter tter.. The strength strength of weak ties. ties. Ame Americ rican an Jour Journal nal of So Sociolo ciology gy , 78: 1360–1380., 1973. Luigi Lui gi Gui Guiso, so, Pa Paola ola Sapienza, Sapienza, and Lui Luigi gi Zin Zingale gales. s. Does Culture Culture Affe Affect ct Eco Economi nomicc Outcomes? The Journal of Economic Perspectives , 20(2):23–48, 2006. Henner Hess. Mafia and Mafiosi: The structur structuree of power . Ashgate, 1973. Francis A.J. Ianni and Elisabeth Reuss-Ianni. A Family Business: Business: Kinsh Kinship ip and So Social cial Control in Organized Crime . Russell Sage Foundation, 1972.
James B. Jacobs and Lauryn P. Gouldin. ter?
Cosa nostra:
1 2 9– 18 9 , 1 9 99 . Crim Cr imee an and d Ju Just stic ice e , 25:pp. 12
The final chap-
ISSN 019 01923234.
URL
http://www.jstor.org/stable/1147609 .
Peter Klerks. The Network Paradigm Applied to Criminal Organisations. Transnational Organised Or ganised Crime: Persp Perspect ectives ives on Glob Global al Secur Security ity , page 97, 2003.
Valdis E. Krebs. Mapping Networks of Terrorist Cells. Connections , 24(3):43–52, 2002. Vimal Kumar and Stergios Skaperdas. Criminal Law and Economics , chapter On The Economics of Organized Crime. Edward Elgar Pub, 2009. Steven Stev en D. Lev Levitt itt and Sudhir Alladi Alladi Venk enkates atesh. h. An eco economi nomicc anal analysi ysiss of a drug drug-se -selli lling ng gang’s finances finances.. The Quarterly Journal of Economics , 115(3):755–789, August 2000. Don Liddick. The Enterprise “Model” of Organized Crime: Assessing Theoretical Propositions. Justice Quarterly , 16(2):403–430, 1999. Salvatore Salv atore Lupo. Quando la Mafia Trov`o l’America . Einaudi, 2009. Peter Maas. The Valachi Papers . Putnam, New York, 1968.
36
Jean Marie McGloin. Pol Policy icy And Int Interve erventi ntion on Consi Consideration derationss of a Net Network work Analysis Analysis of Street Gangs. Criminology & Public Policy , 4(3):607–635, 2005. John C. McWilliams. The Pr Prote otector ctors: s: Harr Harry y J. Ans Anslinger linger and the Feder deral al Bur Bureeau of Narcot Nar cotics, ics, 1930-1962 . Univ of Delaware Press, 1990.
Carloo Mors Carl Morsell elli. i. Care Career er opport opportuni unitie tiess and net netwo workrk-bas based ed pri privil vileges eges in the Cos Cosaa Nos Nostra. tra. Crime, Law and Social Change , 39(4):383–418, 2003.
Mangai Natarajan. Unders Understanding tanding the Structure of a Drug Traffick rafficking ing Organization: Organization: A Conversational Analysis. Crime Prevention Studies , 11:273–298, 2000. Mangai Natarajan. Unders Understanding tanding the Structure of a Large Heroin Distribution Distribution Network: Network: A Quantitative Analysis of Qualitative Data. Journal of Quantitative Criminology , 22 (2):171–192, 2006. Eleonor Ele onoraa Pa Patacc tacchin hinii and Yves Yves Zeno Zenou. u. The strength strength of we weak ak tie tiess in crime. European Economic Review , 52:209–236, 2008.
Alejand Ale jandro ro Portes. Portes. Socia Sociall Capital: Capital: Its Origins Origins and Applicati Applications ons in Moder Modern n Socio Sociolog logy. y. Annual Review of Sociology , 24:1–24, 1998.
Albert R´eka, eka, Hawoong Jeong, and Albert L´ aszl´o Barab´ aszl´ asi. Error and Attack Toleranc asi. Tolerancee of Complex Networks. Nature , 406:378–382, 2000. Peterr Reuter. Disor Pete Disorganiz ganizeed Crime Crime:: the Economics Economics of the Visib Visible le Hand . MIT press, 1983. Peter Reuter. The organization of illegal markets: An economic analysis. 1985. Peter Reuter. Handbook of organized crime in the United States , chapter Research on American Organized Crime. Greenwood Press, Westport, CT, 1994. Jerzy Sarnecki. Sarnecki. Delin Delinquen quentt Net Network workss in Swe Sweden. den. Journal of Quantitative Criminology , 6 (1):31–50, 1990. ISSN 0748-4518. 37
Jerzy Sarnecki. Delinquent Networks: Youth Co-offending in Stockholm . Cambridge Univ Pr, 2001. ISBN 0521802393. Friedri riedrich ch Schneider Schneider and Dominik H. Enste. Shado Shadow w economies: Size, causes, causes, and consequences. Journal of Economic Literature , 38(1):77–114, March 2000. Malcolm K. Sparrow. The Application of Network Analysis to Criminal Intelligence: an Assessment of the Prospects. Social Networks , 13(3):251 – 274, 1991. John R. Wagley agley.. Transnati ransnational onal Organized Crime: Princi Principal pal Threats and US Responses Responses.. Congressional Research Service Report for Congress RL33335, March 2006. Stanley Wasserman and Katherine Faust. Social Network Analysis. Methods and Applications . Cambridge: Cambridge university press edition, 1994.
Phil Williams. Williams. Transnati ransnational onal Criminal Net Network works. s. Netw Networks orks and netwars: the future future of terror, crime, and militancy , page 61, 2001. Primee Minis Minister ter of the Underworld Underworld . George Wolf and Joseph DiMona. Frank Costello: Prim
Morrow, 1974.
38
Figure 1: Record Number One: Joe Bonanno
39
Figure 2: The Bonanno Family. Source: Valachi’s testimony in the McClellan Hearings.
40
Figure 3: Surveillance pictures of Nick Giso in 1980, and Sammy Gravano in 1988
0 0 0 0 0 7
2 . 3 1
3 1
0 0 0 0 0 6
e u 8 l . a 2 v 1 e s u o H 6 − . g 2 o 1 l
e u l a 0 v 0 0 e 0 s u 0 o 5 H 0 0 0 0 0 4 0 0 0 0 0 3
4 . 2 1
0
2
4
6
8
2 . 2 1
10
p_sdeigen
0
2
4
6
8
p_sdeigen
Figure 4: Housing values and centrality Notes: The figures show house values Notes: values by eigenvec eigenvector tor dec deciles iles together together with the 95 percent confidence interval.
41
10
8 0 .
1 . 8 0 .
6 0 . y t i s n 4 e 0 . D
y 6 t i 0 s . n e 4 D 0 .
2 0 .
2 0 . 0
0
20
40 Degree
60
0
80
5 0 .
2 .
4 0 .
5 1 . y t i s n 1 e . D
y 3 t i s 0 . n e 2 D 0 .
20
40 60 Centrality (std.)
80
100
0
20
40 60 Betweenness (std.)
80
100
5 0 .
1 0 . 0
0
40
60 80 Closeness (std.)
0
100
Figure 5: Densities of Centrality Indices
4 .
3 .
3 0 0 .
1 .
3 0 0 .
5 2 0 0 .
8 0 .
5 2 0 0 .
2 0 0 . t h g i e 5 w 1 0 0 .
g n i r o t i 2 . n o m
1 .
0
0
20
40 Degree
60
80
2 0 0 . t h g i e 5 w 1 0 0 .
6 0 . y t i s n e d 4 0 .
1 0 0 .
2 0 .
1 0 0 .
5 0 0 0 .
0
5 0 0 0 .
0
20
40 Degree
60
80
Figure 6: Monitoring, Density (left axes) and Weighting (right axes) Notes: γ equal to 1, 10, and 15. The weighting functions are decreasing, the monitoring ones increasing in degree degree..
42
6 0 0 .
4 0 0 .
2 0 0 .
0
0
20
40 Degree
60
Optimal m mo onitoring we weights: ga gamma=15 gamma=1
80
gamma=2 MC weights
Figure 7: Weighting based on MC and Optimal Monitoring Notes: γ are equal to 1, 2, and 15. MC weights indicate the Markov Chain based weights.
4 .
2
3 .
5 . 1 y t i s n e 1 D
y t i s n 2 e . D
5 .
1 .
0
.5
.6
.7 .8 Probability of success p_0
.9
0
1
0
10
20
Figure 8: Simulations of Degree Notes: τ is equal to 0.5, p and q are set to be the same.
43
30 Connections
40
50
6 4
4 4
e 2 d 4 u t i g n o l 0 4
8 3
6 3
5
10
15
20
latitude
Figure 9: Geographical Distribution of Mafia Surnames. Notes: Each circle represents a zip code. The size of the circles is proportional to the number of US Mafiamembers’ surnames found in today’s Italian phone directory. The plot shows only 20 percent of the distribution of surnames.
44
Degree
Eigenvector
0 2
0 2
5 1 ) . 0 d t 1 s ( y t i l a r t n e C 0
) . 0 d t 1 s ( e e r g 5 e D 0
5 −
0 1 −
20
40
60
80
20
40
Age
60
80
Age
kernel = epanechnikov, degree = 0, bandwidth = 2.66, pwidth = 3.99
kernel = epanechnikov, degree = 0, bandwidth = 2.84, pwidth = 4.25
Closeness
Betweenness 8
0 7
6
0 6 ) . d t s 0 ( 5 s s e n e s 0 o 4 l C
) . d t s ( 4 s s e n n e e 2 w t e B
0 3
0
2 −
0 2
20
40
60
80
20
40
Age
60
80
Age
kernel = epanechnikov, degree = 0, bandwidth = 2.42, pwidth = 3.64
kernel = epanechnikov, degree = 0, bandwidth = 3.53, pwidth = 5.29
Figure 10: Centrality and Age Notes: The figures show Kernel Notes: Kernel-wei -weighte ghted d local linear regression and the corre corresponding sponding 95 percen percentt confidence interval.
45
Degree
Eigenvector
0 6
0 5
0 4
0 4 ) . d t s ( e 0 e 2 r g e D
) . 0 d t 3 s ( y t i l a r 0 t n 2 e C
0
0 1
0
0 2 −
0
5
10
15
0
5
Interaction index
10
15
Interaction index
kernel = epanechnikov, degree = 0, bandwidth = .64, pwidth = .96
kernel = epanechnikov, degree = 0, bandwidth = .74, pwidth = 1.11
Closeness
Betweenness 0 4
0 8
0 3
0 7 ) . d t s ( s s 0 e 6 n e s o l C
) . d t s ( 0 s 2 s e n n e e 0 1 w t e B
0 5
0
0 1 −
0 4
0
5
10
15
0
Interaction index
5
10
15
Interaction index
kernel = epanechnikov, degree = 0, bandwidth = .92, pwidth = 1.38
kernel = epanechnikov, degree = 0, bandwidth = .73, pwidth = 1.09
Figure 11: Centrality and “Tradition” Notes: The figures show Kernel Notes: Kernel-wei -weighte ghted d local linear regression and the corre corresponding sponding 95 percen percentt confidence confid ence interval. interval. The interaction interaction index has been b een truncated at the 95th percen percentile. tile.
46
40−60 percentile range 5 2
0 2
5 1
0 1
5
0
2
4
40/60 sons
40/60 daughters
Figure 12: Centrality, Sons and Daughters Notes: Each line represent the 40-60 percentile range.
47
6
40−60 percentile range 0 2
5 1
0 1
5
0
2
4
6
40/60 crimes
8 40/60 businesses
Figure 13: Centrality, Types of Crimes and of Businesses Notes: Each line represent the 40-60 percentile range.
48
10
49
Table 1: Summary Statistics of Individual Characteristics Variable
Mean
Weighting
5 0
Std. Dev. γ = ∞
Mean
Std. Dev.
Mean
Std. Dev.
Min
Max
Obs
0 0 0 23 5 95
1 1 1 81 6 .25 6. 365
801 801 801 801 790 790
Born in the U.S. Born in Italy (except Sicily) Born in Sicily Age Height in feet Weight in p ounds
0.59 0.19 0.29 52.17 5.61 176
0.49 0. 39 0.45 10.04 0. 20 27
MC sampling γ = 15 Variables related to the PERSON 0. 60 0.49 0.59 0.49 0. 21 0.41 0. 22 0.42 0. 24 0.43 0. 25 0.43 51.58 10.02 51.98 10.41 5. 60 0.20 5. 60 0.20 175 28 175 28
Interaction index Married Divorced Connected wife Numb er of children Fraction of daughters Siblings Extended family memb ers
2.67 0.73 0.05 0.18 1.02 0.49 1.97 1.62
5. 83 0.44 0.22 0.38 1.44 0. 37 2. 11 1. 05
Variables related to the FAMILY 2. 01 4.39 1. 80 3.74 0. 70 0.46 0. 70 0. 46 0.05 0.23 0. 06 0. 24 0. 15 0.36 0. 14 0.35 0.94 1.40 0.97 1.40 0.50 0.38 0. 49 0. 38 1. 93 2.09 1. 87 2. 07 1. 39 0.86 1. 34 0.77
0 0 0 0 0 0 0 1
53.91 1 1 1 8 1 11 6
801 801 801 801 80 801 352 801 801
0. 50 0.24 0.48 9.06 0. 37 1. 70 0. 97
Variables related to the ACTIVITIES 0. 42 0.49 0. 43 0.50 0. 06 0.24 0.06 0.24 0. 61 0.49 0.62 0.49 25.14 9.29 25.19 9. 50 0. 17 0.37 0. 17 0.37 2. 53 1.70 2.52 1. 76 1.01 0.94 1. 03 0.93
0 0 0 8 0 0 0
1 1 1 67 1 9 5
801 88001 801 80 688 801 80 801 801
Resides in NY Resides in NJ Violent crimes Age of first arrest Never arrested Typ es of crimes committed Typ es of businesses
0.43 0.06 0.63 25.02 0.16 2.58 1.07
Table 2: Summary Statistics of Legal Businesses Variable Drugstores Restaurants Fo o d companies Manual laborer Casinos Real estate Imp ort exp ort Car dealer
Mean
Std. Dev.
γ = ∞ 0.18 0.38 0.09 0.31 0.09 0.28 0.07 0.26 0.07 0.25 0.05 0.23 0.05 0.22 0.05 0.22
Mean
Std. Dev.
γ = 15 0.17 0.37 0.09 0.30 0.08 0.28 0.09 0.28 0.05 0.23 0.05 0.22 0.05 0.22 0.05 0.21
Mean
Std. Dev.
MC sampling 0.17 0.37 0.09 0.29 0.08 0.28 0.09 0.29 0.05 0.23 0.06 0.24 0.05 0.22 0.04 0.20
Table 2: Summary Statistics of Legal Businesses Variable Drugstores Restaurants Fo o d companies Manual laborer Casinos Real estate Imp ort exp ort Car dealer
Mean
Std. Dev.
Mean
γ = ∞ 0.18 0.38 0.09 0.31 0.09 0.28 0.07 0.26 0.07 0.25 0.05 0.23 0.05 0.22 0.05 0.22
Std. Dev.
γ = 15 0.17 0.37 0.09 0.30 0.08 0.28 0.09 0.28 0.05 0.23 0.05 0.22 0.05 0.22 0.05 0.21
Mean
Std. Dev.
MC sampling 0.17 0.37 0.09 0.29 0.08 0.28 0.09 0.29 0.05 0.23 0.06 0.24 0.05 0.22 0.04 0.20
Table 3: Summary Statistics of Crimes Variable Drug offenses Robb ery Murder Weap on offenses Simple assault Larceny Burglary Gambling Liquor offenses Extortion Counterfeiting
Mean
Std. Dev.
Mean
γ = ∞ 0.43 0.50 0.26 0.44 0.23 0.42 0.22 0.42 0.21 0.41 0.20 0.40 0.13 0.34 0.13 0.33 0.13 0.34 0.07 0.25 0.07 0.25
Std. Dev.
γ = 15 0.46 0.50 0.27 0.44 0.19 0.40 0.21 0.41 0.21 0.41 0.20 0.40 0.13 0.34 0.13 0.34 0.12 0.32 0.06 0.23 0.07 0.26
51
Mean
Std. Dev.
MC sampling 0.44 0.50 0.27 0.44 0.20 0.40 0.21 0.41 0.22 0.42 0.20 0.40 0.14 0.34 0.14 0.35 0.11 0.31 0.06 0.24 0.08 0.27
Table 4: List of Criminals with the Highest and the Lowest Degree Lastname Lucania Ormento Accardo Accardo Genovese Genovese Copp ola Copp ola Copp ola Strollo Profaci Profaci Santoro Vitale Vitale
Name Salvatore John Settimo Antonio Vito Michael Michael Frank Stephen Antonio Frank Joseph Salvatore Vito Salvatore
Degr De greee
Top
High Hi gh
Apal Ap alac achi hin n Rank (Family)
Criminals with the highest 15 degrees 71 1 1 0 Boss 71 0 2 1 Cap oregime (Lucchese) 64 1 2 0 Cap oregime (Lucchese) 64 0 1 0 Boss 55 1 0 1 Underb oss (Luciano) later Boss 55 0 1 1 Boss 54 2 1 0 Cap oregime (Genovese) 54 0 5 0 Dep orted to Italy 54 0 1 0 Soldier (Maggadino) 47 2 2 0 Cap oregime (Genovese) 44 1 1 0 Boss 44 1 2 1 Boss (son of Frank) 42 0 0 0 Underb oss (Lucchese) 40 1 1 0 Boss in Sicily 40 1 0 0 Underb oss (Bonanno)
Criminals with the lowest 15 degrees Castorina Vincent Vi 1 0 0 Kornhauser Max 1 0 2 Bibb o Nicholas 1 0 0 Virusso Santo 1 0 1 Mandala Nicholas 1 0 0 Rob erto Dominick 1 0 1 Simoni Pierre 1 0 0 Candelmo John 1 0 1 Colomb o Frank 1 0 0 Bongiorno Frank 1 0 1 Amari Philip 1 0 1 Peloso Antonio 1 0 0 Labarbara Joseph 1 0 0 Pine Grace 1 1 1 Valle Alarico 1 1 0
52
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
Table 5: Summary Statistics of Network-related Variables Variable
Mean
S td. Dev. St
Mean
γ = ∞
Degree Degree (std.) Centrality (std.) Closeness (std.) Betweenness (std.) Top ranked Top rank citations
11.14 14.49 12.57 52.53 5.10 0.55 0.85 0.
S td. Dev. St
Mean
S td. Dev. St
MC sampling Measures of centrality 7.28 7.37 6. 46 5.79 8.98 10.54 7.80 8.27 7.82 10.70 6.93 8.57 45.52 14.86 44.39 14.33 2.78 6.59 2. 01 4. 45 0.51 0.50 0.52 0.50 0. 75 0.94 0.77 0.95
Mi n
Max Ma
1 0 0 0 0 0 0
7711 100 100 100 100 1 7
γ = 15
9.55 13.64 14.16 15.25 9. 46 0.50 1.02
Table 6: Average Centrality by Rank ran ank k
N. ob obss.
degr greee
3 36 2 32 79 13 24 1 18
12.99 14.44 19.11 24.16 22.24 21.29
unknown soldier captain consigliere underb oss b oss
eige gen nvector cl clos oseene nesss 9.90 11.44 17.35 24.02 16.61 17.20
71.11 72.19 75.52 78.65 75.89 75.33
bettweenn be nneess 4.06 3.80 6.48 6.76 9.88 8.50
Table 7: Housing value regressions
log-centrality index
(1)
(2)
(3)
0.157** **** (0.044)
0.0966** 0.09 (0.048)
0.108* 8**** (0.028)
(4) (5) log House value
0 . 24 8** (0.1 (0 .104 04))
Resides in NJ Other Xs State FE Observations R-squared
no no 610 0. 04 0
0. 8 99** * (0.104) 0.385** (0.154) yes no 6 10 0. 222
no yes 6 09 0. 3 73
(7)
(8)
0 . 22 0 * * (0.0 (0 .094 94)) 0. 9 30 ** * (0.104) 0. 447 *** (0.135) yes no 6 10 0. 2 17
0. 09 6 (0.0 (0 .093 93))
0 . 15 7* (0.0 (0 .089 89))
no yes 609 0. 359
yes yes 60 9 0.394 0.
0.085* 0.0 ( 0. 043 )
boss highest most head top
Resides in NY
(6)
yes yes 6 09 0. 399
no no 6 10 0. 0 10
Notes: The additional regressors (“Other Xs”) are the extended family members, being born in Sicily, or in the rest of Italy, age, age squared, the interaction index, height, weight, marital status, whether the wife is connected, the number of children, the fraction of daughters, the number of siblings, violent crimes, age at first arrest, never arrested, the types of crime committed, and the types of businesses. Clustered (by family) standard errors in parentheses: *** p <0.01, ** p<0.05, * p<0.1.
53
Table 8: Positive Assortativity Among Crime and Business Types Crimes
β β SD(β β ) Murder 0.440 0.083 Robbery 0.478 0.091 Simple assault 0.339 0.087 Burglary 0.232 0. 0 .099 Larceny 0.068 0.094 Counterfeiting 0.361 0.148 Drug offenses 0.939 0.053 Gambling 0.608 0.117 Liq Li quor off offeenses -0.001 0.072 Wea eapo pons ns off offen ensses 0. 0.16 1600 0. 0.08 0866 Extortion 0.403 0.135
Businesses β β SD(β β ) Restaurants 0.158 0.095 Drugstores 0.184 0.098 Foo ood d companies 0.447 0.111 Real estate 0.310 0.134 Imp ort exp ort 0.403 0. 0.155 Manual labo borrer 0.215 0.129 Casinos 0.348 0.111
Notes: Estimated coefficients and clustered standard errors by surnames from a regression of crime type dummies on the fraction of associates who perpetrated the same crime type. I restrict the data to businesses held and crimes perpetrated by at least five percent of the sample.
54
Table 9: The Determinants of the Eigenvector Index (1) Weighting: γ Ext xteend nded ed fa fami milly me mem mbe bers rs Born in Italy (except Sicily) Born in Sicily Age Age squared/100 Interaction index Height in feet Weight in pounds
∞ 4.672* 4.67 2*** ** (0.905) -1 . 0 4 2 (1.128) 2. 5 78* (1.466) 0. 89 9 *** (0.304) - 0. 765 ** (0.300) 2. 5 12* (1.407) 2 . 049 (2.580) 0 . 001 (0.020)
Married Divorced Connected wife Number of children Fraction of daughters Siblings Resides in NY Resides in NJ Violent crimes Age at first arrest Never arrested Types of crime committed Types of businesses Observations R-squared Mean value:
8 01 0 . 195 1 2. 5 7
(2)
(3) (4) Eigenvalue index MC 15 ∞ 3.54 3. 541* 1*** ** 4. 4.08 085* 5*** ** 5.04 5. 041* 1*** ** (0.817) (1.084) (0.764) - 0 . 872 -0 . 6 8 6 -0 . 5 9 9 (0.653) (0.728) (1.056) 1. 82 7 ** 2 . 584* * 3 . 38 7** (0.822) (1.034) (1.381) 0. 341 ** 0.338 0. 0 . 90 9*** (0.169) (0.218) (0.315) - 0. 0.289* -0.257 - 0. 768 ** (0.168) (0.214) (0.311) 2. 487 ** 3 . 188* * 2. 3 88 * (1.144) (1.333) (1.316) 2. 71 5 * 3 . 553* 2.534 (1.596) (1.864) (2.423) 0. 00 4 0. 002 -0.004 (0.011) (0.013) (0.018) -0.737 (1.057) -0.507 (2.369) 4. 50 5** * (1.546) 0. 80 8** (0.347) 1.778 (1.903) -0.115 (0.218) 5 . 29 7*** (1.123) 2 . 7 3 4* (1.650) 1.337 (0.931) - 0. 0 57 * (0.034) - 2. 176 * (1.199) 0 . 7 7 3 ** (0.299) 1 . 91 5*** (0.527) 80 1 80 1 8 01 0. 17 1 0. 194 0.287 6. 9 2 7 . 81 1 2. 57
(5)
(6)
MC 3.991* 3.99 1*** ** (0.737) - 0. 2 64 (0.663) 2 . 183 ** * 2. (0.814) 0. 315 * (0.178) - 0. 2 43 (0.177) 2. 3 63* * (1.099) 2. 1 37 (1.483) 0. 0 04 (0.009) 0. 288 (0.596) - 1 . 137 (1.400) 2. 925* ** (0.861) 0 . 150 (0.212) 1. 022 (1.087) - 0 . 093 (0.132) 3. 485* ** (0.626) 2. 3 76 * (1.346) - 0. 036 (0.492) - 0. 03 2 (0.023) - 0 . 741 (0.794) 0. 2 19 (0.203) 0. 666 * * (0.283) 80 1 0. 239 6 . 92
15 4.512* 4.51 2*** ** (0.928) -0.022 (0.739) 2 . 9 0 4*** (1.021) 0.425* 0. (0.228) -0.332 (0.225) 3.081** (1.261) 2. 963 * (1.727) 0. 002 (0.011) -0.139 (0.737) -0.903 (1.574) 3 . 6 7 9*** (1.096) 0. 450* (0.253) 0. 585 (1.219) -0.161 (0.152) 4. 0 27* ** (0.753) 2.550** (1.257) 0 . 466 (0.578) -0.041* (0.025) -1.297 (0.840) 0 . 334 (0.245) 1 . 1 7 8*** (0.369) 8 01 0. 26 9 7. 81
Notes: The regressions include also missing dummies for year of birth, height, and weight. Clustered (by family) standard errors in parentheses: *** p<0.01, ** p<0.05, * p<0.1.
55
Table 10: The Determinants of Importance Indices ( 1) Degree Extended family membe berrs
4.5966*** 4.59 (0.840) Born in Italy (except Sicily) -0.259 (0.748) Born in Sicily 1.572** (0.789) Age 0.429** (0.176) Age squared/100 -0.378** (0.172) Interaction index 2.402** (1.095) Height in feet 2.661* (1.593) Weight in pounds 0.006 (0.011) Married 1.051* (0.630) Divorced -1.441 (1.313) Connected wife 2.848*** (0.790) Number of children 0.254 (0.209) Fraction of daughters 0.534 (1.007) Siblings 0.033 (0.130) Resides in NY 1.182** (0.588) Resides in NJ -0.196 (1.190) Violent crimes -0.341 (0.491) Age at first arrest 0.011 (0.025) Never arrested 0.087 (0.885) Types of crime committed 0.328 (0.207) Types of businesses 0.405 (0.250) Observations R-squared
801 0.282
( 2) (3) Eigenvalue Total Residual
( 4) ( 5) Closeness Total Residual
3.991** **** -0.248 (0.737) (0.344) -0.264 -0.025 (0.663) (0.415) 2.183*** 0.732* (0.814) (0.445) 0.315* -0.081 -0 (0.178) (0.118) -0.243 0.105 (0.177) (0.115) 2.363** 0.148 (1.099) (0.588) 2.137 -0.317 (1.483) (0.909) 0.004 -0.002 (0.009) (0.006) 0.288 -0.682 -0 (0.596) (0.424) -1.137 0 .193 0. (1.400) (0.687) 2.925*** 0.298 (0.861) (0.468) 0.150 -0.084 (0.212) (0.113) 1.022 0.529 (1.087) (0.589) -0.093 -0.123* (0.132) (0.072) 3.485*** 2.395*** (0.626) (0.378) 2.376* 2. 2.556*** (1.346) (0.669) -0.036 0.278 (0.492) (0.341) -0.032 -0.042** (0.023) (0.017) -0.741 -0.821 (0.794) (0.580) 0.219 -0.084 (0.203) (0.109) 0.666** 0.293* (0.283) (0.168)
6.236** **** 2.201** **** (1.194) (0.407) -2.427 -0.053 (1.640) (0.290) 2.705 0.668** (1.710) (0.312) 1.294*** 0.221*** (0.456) (0.080) -1.173*** -0.208*** (0.447) (0.078) 3.868*** 2.455*** (1.379) (0.838) 5.111 0.455 (3.410) (0.614) 0.017 0.000 (0.029) (0.004) 0.694 -0.019 (1.371) (0.253) -4.229 -0.222 (3.291) (0.448) 3.887** 1 .211*** 1. (1.607) (0.318) 0.207 0.176** (0.520) (0.076) 1.498 -0.180 (2.639) (0.373) 0.249 0.048 (0.312) (0.051) 6.704*** 0.144 (1.334) (0.227) 7.541** 0.169 (3.434) (0.463) -0.530 -0.175 (1.161) (0.204) 0.001 0.002 (0.055) (0.010) -2.122 0.249 (2.009) (0.361) -0.103 0.105 (0.482) (0.073) 1.130* 0.193* (0.597) (0.102)
801 0.239
801 00..079
8 01 0.271
8 01 0.272
(6) (7) Betweenness Total Residual 2.230*** (0.835) -2.201 (1.371) 1.335 (1.265) 0.920** (0.391) -0.843** (0.377) 1.774* (1.045) 2.791 (2.791) 0.012 (0.025) -0.222 (1.170) -2.973 (2.470) 1.405 (1.203) -0.014 (0.385) 1.033 (2.009) 0.220 (0.240) 5.674*** (1.033) 7.712*** (2.626) -0.234 (0.958) -0.009 (0.049) -2.199 (1.806) -0.389 (0.370) 0.777 (0.480)
0.031 (0.042) 0.162 (0.107) -0.046 (0.097) -0.002 (0.030) 0.019 (0.029) -0.073 (0.070) -0.020 (0.202) 0.002 (0.002) -0.125 (0.099) 0.190 0. (0.193) 0.206* (0.117) -0.028 (0.028) 0.218 (0.142) 0.041** (0.020) 0.104 (0.087) 0.274 (0.183) 0.126 (0.087) -0.001 (0.003) 0.150 (0.141) -0.002 (0.029) 0.052 (0.041)
801 0.178
801 00..101
Notes: All regressions use weights based on γ = 15. The regressions include also missing dummies for year of birth, height, and weight. The leader variable counts the number of words that describe mobsters as leaders. Clustered (by family) standard errors in parentheses: *** p <0.01, ** p<0.05, * p<0.1
56
Table 11: Robustness Checks ( 1) (2) Eigenvalue Weighting: γ Extended fam amiily membe berrs Born in Italy (except Sicily) Born in Sicily Age Age squared/100 Interaction index Height in feet Weight in pounds Married Divorced Connected wife Number of children Fraction of daughters Siblings Resides in NY Resides in NJ Violent crimes Age at first arrest Never arrested Types of crime committed Types of businesses Observations R-squared
(3) (4) Leadership
∞
15
∞
15
5.041* 1**** (0.764) - 0. 59 9 (1.056) 3. 387 ** (1.381) 0. 9 09 *** (0.315) - 0. 768 * * (0.311) 2. 38 8* (1.316) 2. 5 34 (2.423) - 0. 004 (0.018) - 0. 737 (1.057) - 0. 507 (2.369) 4. 505 *** (1.546) 0. 8 08 ** (0.347) 1. 7 78 (1.903) - 0. 115 (0.218) 5. 29 7 ** * (1.123) 2 . 73 4* (1.650) 1. 3 3 7 (0.931) - 0 . 057* (0.034) - 2. 1 76* (1.199) 0. 773 ** (0.299) 1. 915 ** * (0.527)
3.991*** 3.9 (0.737) -0.264 (0.663) 2 . 183* * * (0.814) 0 . 315* (0.178) - 0. 2 43 (0.177) 2. 363 * * (1.099) 2. 137 (1.483) 0 . 004 (0.009) 0 . 288 (0.596) - 1. 13 7 (1.400) 2 . 925* * * (0.861) 0 . 150 (0.212) 1. 022 (1.087) - 0. 093 (0.132) 3 . 485** * (0.626) 2 . 376* (1.346) - 0. 0 36 (0.492) - 0. 032 (0.023) - 0. 7 41 (0.794) 0. 21 9 (0.203) 0. 66 6* * (0.283)
-0.009 (0.037) 0. 0 39 (0.093) 0. 1 53 (0.099) 0. 0 15 (0.027) -0.002 (0.026) -0.056 (0.058) -0.104 (0.206) 0. 0 01 (0.001) -0.051 (0.087) 0. 0 83 (0.163) 0. 1 07 (0.104) 0. 0 12 (0.029) 0. 1 67 (0.142) 0. 0 37* (0.019) 0. 16 6 ** (0.084) 0. 3 23* (0.189) 0. 0 84 (0.082) -0.002 (0.003) -0.053 (0.112) 0. 0 14 (0.024) 0. 08 7** (0.040)
0. 031 (0.042) 0. 16 2 (0.107) -0.046 (0.097) -0.002 (0.030) 0. 01 9 (0.029) -0.073 (0.070) -0.020 (0.202) 0. 002 (0.002) -0.125 (0.099) 0. 1 90 (0.193) 0. 206 * (0.117) -0.028 (0.028) 0. 21 8 (0.142) 0 . 04 1** (0.020) 0. 10 4 (0.087) 0. 27 4 (0.183) 0. 126 (0.087) -0 . 0 0 1 (0.003) 0. 15 0 (0.141) -0 . 0 0 2 (0.029) 0. 05 2 (0.041)
8 01 0. 2 87
80 1 00..239
8 01 0. 0 71
801 80 0. 1 01
Notes: The dependent Notes: dependent va variable riable “Leadership” “Leadership” counts counts the num number ber of times the words “boss,” “highest,” “most,” “head,” ‘top,” “high,” “influential,” “important,” “leader,” “leading,” “powerful,” and “represent “repre senting” ing” are cited. The regre regressions ssions include include also missing dummi dummies es 57 for year of birth, height, and weight. Clustered (by family) standard errors in parentheses: *** p <0.01, ** p<0.05, * p<0.1.
Tabl ablee 12: Regr Regress ession ionss of Indegree with Random Truncation of Half of the Sample
Extended family members Born in Italy (except Sicily) Born in Sicily Age Age squared/100 Interaction index Height in feet Weight in pounds Married Divorced Connected wife Number of children Fraction of daughters Siblings Resides in NY Resides in NJ Violent crimes Age at first arrest Never arrested Types of crime committed Types of businesses
β
se(beta)
5. 54 - 0. 44 2. 5 8 1. 11 - 1. 00 2. 77 3 . 09 0. 00 0. 04 - 1. 54 4. 78 0. 81 0. 53 0. 12 2. 11 0. 69 0. 69 - 0. 0 2 - 1. 70 0. 6 4 1. 34
0.86 0. 9 7 1. 2 9 0.31 0.30 1.26 2.33 0.02 0.99 1.96 1.50 0.33 1.58 0. 1 9 0. 9 5 1. 8 9 0. 8 6 0.03 1.12 0.27 0.47
simulated aver av erage age me medi dian an 6.03 5. 9 2 -0 . 2 9 - 0. 26 2. 9 7 3. 0 8 1.15 1. 1 4 -1 . 0 3 - 1. 02 3.02 2. 7 6 3.07 3. 2 4 0.01 0. 0 0 0 .63 0. 0. 5 4 -0.58 - 0. 60 4.28 4. 0 2 0.82 0. 8 0 0.41 0. 4 5 0 .0 9 0. 0. 0 9 2. 5 3 2. 5 6 0. 7 1 0. 6 5 0. 6 3 0. 5 6 -0.02 - 0. 0 2 -1.78 - 1 . 70 0.67 0. 6 7 1.48 1. 5 2
Notes: The first two columns show the coefficients and the standard errors based on the whole sample. Columns 3 and 4 show the mean and the median coefficients of 500 regressions based on randomly truncated samples.
58