ACES Panel Paper
Membership Has Its Privileges: Old and New Mafia Organizations Maria Minniti Skidmore College
What makes a country, or a geographical area, more likely to develop a mafia? At the beginning we find a large-scale shock to the social system that pushes it ffom a low-crime equilibrium without a mafia toward a high-crime equilibrium with a mafia. Following the shock, the necessary condition for the development o f a mafia is the existence of a dynamic, which I call social contagion, similar to a technological lock-in. Indeed, the outbreak of a majulike organization creates a self-reinforcing process. Consequently, anti-mafia policies will be successjid only ifable to push the system back toward a lowcrime mafia-free equilibrium. Unfortunately, however, if being a member of the mafia begins to emerge as the dominant choice, elimination of the organization becomes progressively more dificult as the necessary changes in relative returns to membership increase at an increasing rate.
The current literature contains conflicting definitions and explanations of mafias. (See, for example, Gambetta 1992, Jennings 1984, and Shelling 1967.) Some researchers define mafias as informal governments that provide police and judicial services. Others claim that the essence of a mafia is the establishment of monopoly power by erecting barriers to entry. Still others emphasize the relationship between a mafia and the presences of high returns to be extracted from illegal markets. Considerable complementarity exists among these views, all of them illuminate aspects of mafia-like organizations. For the purposes of this paper, however, the central properties of a mafia are the large number of people involved, and the organization's strongly self-perpetuating character. The key observation is that, for the mafia to evolve and establish itself, large numbers of adherents are necessary. Indeed, to prosper, the mafia relies on the collective outcome of many individual contributions, no matter how obtained.
Comparative Economic Studies
Vol. 37, No. 4, Winter 1995
Mafia-like organizations are not limited to Southern Italy. Although it is widely believed that, compatible with historical circumstances, similar institutions should deliver similar economic and social results, when it comes to mafia-like organizations, cross-country observations show that they may originate and develop from very different environments. We also observe that, given similar institutional settings, some societies develop a mafia whereas others do not. Is it then possible to identify what makes a country, or a geographical area, more like to develop a mafia? My claim is that, unfortunately, it is not. The development of a mafia-like organization is, in fact, the result of two phenomena. First, the necessary condition for the onset of a mafia is a large-scale shock to the social system that pushes it from a low-crime equilibrium without a mafia toward a high-crime equilibrium with a mafia. Such a shock may orginate from the most disparate events. Second, the necessary condition for the development of a mafia is the existence of a network externality created by those belonging to the organization over the rest of the community. Such social dynamics, which I call social contagion, is characteristic of the mafia itself and does not depend on social or institutional characteristics of the economy. Indeed, the outbreak of a mafia-like organization creates a network externality such that its spreading thereafter becomes a self-reinforcing process. This phenomenon may be described as the realization of a path-dependent social process: as the mafia spreads, membership in its ranks becomes more attractive; that is, it exhibits increasing returns to adoption. My argument contributes to explain why, once established, all mafia organizations end up looking similar and are difficult to eradicate. If the mafia reaches a critical size with respect to total population, society becomes locked into a technology of crime, and the choices of new individuals entering the economy become path-dependent. I will show that if honest work and membership in the mafia are the only two alternative income producing activities, then: the difference between the returns to these two activities, called the relative-return to membership, may be described as a technology exhibiting increasing returns to adoption; and, the elimination of the mafia becomes progressively more difficult as being part of the organization begins to emerge as the dominant choice. From the latter, it follows that the adjustment necessary to shift the system back toward a mafia-free environment increases at an increasing rate. Within the context of the literature on mafia and organized crime, this paper indicates a possible explanation for the historically experienced difficulties of eradicating mafia-like organizations and suggests some necessary features of potentially successful anti-mafia policies.
OLD AND NEW MAFIA ORGANIZATIONS
Membership Has Its Privileges In recent years Case and Katz (1991), Crane (1991), Granovetter and Soong (1983), and Miyao (1978) have pointed out the importance of individuals' interdependence in generating alternative patterns of social behavior. When studying a person's choice, we should not abstact from the fact that individuals face different social environments. Like it or not, we make the choices we make and behave the way we do because of the specific knowledge of the world we have. We may think of individuals and social environment in the same way we think of individual producers in a competitive market. In spite of contributing directly to the size of the market, each producer is still a price taker. In this paper, however, I only consider those aspects of human interactions that generate network effects; that is, those that create externalities with self-reinforcing characteristics. I call this particular form of interdependence social contagion. Social contagion may be effectively described by using models of technology adoption similar to those used to study engineering or industrial organization problems. In these models, the existence of self-reinforcing properties enables relatively insignificant circumstances to become important by allowing them to push the system toward one outcome rather than another. Similarly, through social contagion, the behavior of each individual contributes to create a network externality whose perception, in turn, affects other individuals' decisions. Mafia-like organizations are particularly well suited to be described as self-reinforcing phenomena since it is almost impossible to deny the network externalities-the social contagion that the mafia imposes on a community. Any individual living in a community where the mafia exists, in each period of time, faces the choice between working honestly or becoming a member of the organization. Such decision, of course, is determined by several personal and social variables. Among the social ones, the power exercised by the mafia itself is, clearly, an especially important one. Fear, intimidation, lack or difficulty of alternatives, are all considerations that may push an individual toward choosing to cooperate with the mafia rather than to remain honest. In this sense, the existence of the mafia by itself influences people toward a propensity for crime. Traditionally, researchers have studied crime, in all of its forms, by focussing primarily on cost-benefit analyses. (Classic examples of such literature are Becker and Landes 1974, Brower and Ehrlich 1987, and Pyle 1983.) An individual's decision to adhere to the mafia is, indeed, based on the same principle. Prospects of employment, education, family income
'
and structure, together with personal traits like time preferences or disposition to cruelty, all influence a person's propensity to engage in criminal activities. These variables constitute each person's initial endowment, and since they vary across individuals, the population is heterogeneous with different individuals facing different opportunity costs of committing crime. After correcting for these initial differences, however, we would expect the distribution of criminals within each community to be similar across communities. But the mafia is always very concentrated in specific geographical areas. It is not uncommon, for example, to see the entire population of a small Southern Italian town attending the funeral of some well-known mafia criminal. Should we then conclude that, in those areas, the concentration of people more predisposed to crime is higher? The answer is, almost certainly, no. There is no scientific evidence nor logical reason indicating that the number of potential criminals in some Southern towns is higher than in the rest of Italy. Nor is there any reason to believe that the concentration of crime-prone individuals in the districts of Cali and Medallin is higher than in the rest of Colombia. The existence of the mafia itself, thus, is an important factor in an individual's decision to become or not become a criminal. If there is a network externality, ex-ante knowledge of individuals' preferences, though necessary, is not sufficient to anticipate social outcomes. In particular, if membership in the mafia exhibits increasing returns to adoption, the resulting contagion effect may push new individuals into replicating the choices of their predecessors in spite of their own initial preferences. Let honest work and being in the mafia be the only two available incomeproducing activities, and let the relative size of the mafia be a measure of its influence on individuals' lives. Further, let the relative size of the mafia be measured by the ratio of individuals who chose to be in the mafia to the total population. Taking into account all costs and benefits associated with both activities, individuals always choose the one with the highest return. Returns, however, change as the relative size of the mafia changes. For example, as the relative size of the mafia increases, returns to honest work are immediately penalized by monetary extortion and because of increased personal risk and uncertainty. The returns to being in the mafia, on the other hand, begin declining only after the organization has reached a critical size; that is, when competition among criminals begins to dissipate their rents3 Thus, the returns to both activities are decreasing with respect to the relative size of the mafia, but the rates of decline differ. Because of this asymmetry, and because there are only two activities, each individual's choice is determined by the difference between the returns to
OLD AND NEW MAFIA ORGANIZATIONS
35
crime and the returns to honest work; that is, by what I call the relativereturns-to-membership in the mafia. For each individual, the relative-returnto-membership is a function of a set of personal characteristics, the individual's initial endowment, and of the relative size of the mafia itself. Since the returns to being in the mafia are penalized more slowly than those to honest work, the larger is the relative size of the mafia with respect to total population, the higher is the incentive to become a member rather than remain an honest worker, independent of initial personal characteristics. In this sense, the mafia is socially contagious, and the more widespread, the higher is the probability that a new person entering the community will choose criminal activities over honest work. Consistently with everyday observations, the existence of a self-reinforcing externality partly explains why the crime rate, once elevated, is hard to reduce and why to reduce it, or simply prevent it from rising, may require ever larger fractions of income. In addition, the observation that initial preferences are not the only determinant of an individual's choice sheds some light on why communities with initially similar characteristics may find themselves locked into different crime patterns.
A Model of Mafia Memberships A simple model describing how and why mafia membership increases may be built by using a class of dynamic systems known as non-linear pathdependent stochastic processes. These systems have been used in economics since the early 1980s to study technology adoptions (Arthur 1989). In nonlinear path-dependent processes, for any random sequence of events selected from n available types, the probability of each event being repeated is a function of its share of previous realizations. This means that the accumulation early on of some specific events pushes the process toward one among all alternative sequences of events and, possibly, locks the future dynamics of the process into that specific pattern.4
The Return Functions Consider a community where income is obtained only by honest work or by becoming a member of the mafia. In each period of time, each agent j comes into the system and is faced with the choice between the two available income-producing activities: honest work (w) or membership in the mafia (m). The returns realized by each agent, after choosing one of the two activities, depend on a set of personal characteristics; that is, on her
initial endowment and on the relative size of the mafia at the moment of choice. Let the relative size of the mafia be described by cN=M/N, where CN measures the proportion of individuals M who chose to be in the mafia after N individuals have made their choice. For each c,,, agents always choose the technology whose returns are higher.5 For each agent, j, let
describe the returns to being in the mafia as a function of j's personal characteristics and of the relative size of the mafia. And let
describe the returns to honest work as a function of personal characteristics and of CN. Since drydc < 0, d2r?dc2 < 0, dri"/dc< 0, and d2ri"/dc2< 0, both activities exhibit decreasing returns with respect to the relative size of the mafia, with rates of decline increasing in c . ~ For each period of time, let the community be formed by individuals uniformly distributed on the closed unit interval. Let c ~ j m - a p a ~Assume a, # a j for all jf i , so that individuals are heterogeneous and aj influences their choices between the mafia and honest work. Just as individuals are uniformly distributed over the interval [O, 11, the aj values are uniformly distributed over the closed interval [ao,al],where
and
(3)
Note that, unless all agents are assumed to be criminals under all circumstances, a. should not be positive. Similarly, unless all agents are assumed to be always honest, a , should not be negative. From (1) and (2), it is possible to derive, for each individualj, a relativereturn function of the type
Equation (4) shows that, although heterogeneous in their personal
OLD AND NEW MAFIA ORGANIZATIONS
37
characteristics, individuals respond similarly to the presence of the mafia. This similarity is reflected in the fact that each individual's relative-return function, rj, is a vertical displacement from the common function$ Thus, each person's relative-return function is
where r j = r F r y i s derived from equations (1) and (2) and measures the difference between the returns from the two activities for individualj, that is, the relative-return to membership in the mafia. The shape off reflects increasing returns to adoption with respect to the relative size of the mafia and implies that, the more criminals are around, the higher is r,, although its marginal rate is variable. Figure 1 shows the relative-return-to-membership functions for some representative individuals. For individual i, for example, between 0 and A, marginal returns increase at an increasing rate, after the threshold point A, however, they begin decreasing at a decreasing rate. It is important to notice that the choice of most individuals depends on how widespread the mafia is and that different concentration of mafia may mean different choices. Also, consistently with observable human behavior, some individuals are assumed to have a natural preference for honest work and others a natural preference for crime; that is, regardless of the crime rate, some always choose to be in the mafia while others always choose honest work. In Figure 1 such "limit" types are represented by relative-return functions that never cross the horizontal axis.
The Dynamics of the Model Let us now imagine a community with a number of potential newcomers. Each newcomer, j, chooses one income-producing activity. In my model of choice between honest work and the mafia, the realized proportions of each activity are summarized by the number of individuals who chose to be in the mafia.' Thus, a function q(c) may be derived. Such a function maps the relative size of the mafia into the probability that individuals newly entering the community will also choose the mafia over honest work. From (3) and ( 5 ) , such probability is given by
MlNNlTl
FIGURE 1
This can be seen by recalling that the probability of the next agent choosing the mafia is a function of the proportion of agents for whom, given the current relative size of the mafia rj> 0. The function q(c) maintains the cubic shape assumed for the relative-return functions of each individual. Indeed, equation (6) is simply a linear transformation ofJlc). As long as we are willing to assume that all relative-return functions have the same shape, and that agents and their characteristics are uniformly distributed, q will be a linear transformation of the average rj. For any period of time, new individuals who prefer the mafia and new individuals who prefer honest work enter the community with probabilities
OLD AND NEW MAFIA ORGANIZATIONS
39
that are a function of the proportions already realized of their respective preferred activity. The value c,, represents the relative size of the mafia after n -1 agents have made their choice. Let q,, be a continuous function mapping proportions into probabilities at time t. When the n-th agent is added to the community, she chooses the mafia with probability qn(cn).The action is repeated over and over. The key property is that c,, tends, with probability one, to a limit random value c selected from a finite set of possible values. According to Arthur, Ermoliev and Kaniovski's theorem, the set of possible values includes all, and only, the stable fixed points of q. A closer look at the dynamics of the system helps to understand how the process behaves as more and more agents enter the community. l o At time t, the size of the population is N and the number of people in the mafia is M@ As one more person enters the community, the population is (N+n-I) and the number of individuals in the mafia is M,,. Thus, cn=M,,/(N+n-1) measures how widespread the mafia is when individual n makes her choice. When the n + 1 individual is ready to make her choice,
Since individuals have only two alternatives, to become a member of the mafia or to be honest, the size of the mafia is a random variable given by
N+n
I +---N+n
N+n
0 N+n
cn+1=
with probability q(c) with probability 1-q(c)
that is
with probability q(c)
(7) with probability 1-q(c).
Thus, the expected value of the relative size of the mafia is given by
Equation (8) may be rewritten as
Equation (9) describes the deterministic portion of the dynamics of this system. The stochastic portion, instead, is described with the random variable
M c )=
with probability q(c) with probability 1 -q(c).
From equation (7) and (9), it follows that
The stochastic portion of the system's dynamics is given by the third term on the right of equation (lo), and its expected value is zero. Notice that when q(c) % cn,E(c,+ ,) S c,. This shows that the system is attracted to the stable fixed points of q(c). Figure 2 shows the q function. A, B,and C, are fixed points. But, while A and C are stable, B is not. When the q function is approaching point A, the probability of the next agent choosing to become a member is higher than the relative size of the mafia, the latter tends to increase. Between points A and B, the probability of the next agent choosing to become a member is lower than the relative size of the mafia, the latter tends to decrease. Between B and C, the probability of the next agent choosing to be in the mafia is higher than its relative size and the mafia tends to increase. Between C and 1, the probability of the next agent choosing the mafia is lower than the relative size of the mafia and the latter tends to decrease. The
OLD AND NEW MAFIA ORGANIZATIONS
FIGURE 2
Relative Size of the Mafia
system will settle at either A or C. When there are multiple stable points, which one is chosen depends on the accumulation of random events that occur as the stochastic process develops. In other words, where the community settles, whether at C, at a high crime level with a mafia, or at A, at a low crime level without a mafia, depends on the specific sequence of observed choices. In the absence of network externalities, knowledge of each individual's personal characteristics is sufficient to determine her choice between the mafia and honest work. Regardless of the level of crime, the choice is known a priori and the collective outcome is simply the sum of all individual choices. Since relative-returns to membership depend also on how widespread the mafia already is, however, this is not the case in this model. If an initial random shock triggers a sequence of choices in favor of the mafia, membership becomes more appealing to a wider proportion of
potential adopters and, eventually, being in the mafia will emerge as the dominant activity. Different chance events, however, may cause honest work to prevail. This dynamic behavior explains partly why, among communities with initially similar characteristics, some develop a mafia while others do not and why mafia-like organizations tend to be geographically very concentrated. Properties of the Model A closer look at the dynamic properties of the model offers several interesting insights on the progressive development of alternative crime patterns. " First, the process is non-predictable. Predictability would imply the possibility of forecasting exactly what crime rate will emerge in the community. But, by allowing multiple equilibria to exist, the model shows that ex-ante knowledge of individuals' relative-return functions, though necessary, is not sufficient to anticipate which equilibrium will prevail. This means, in particular, that knowledge of people's preferences between honest work and being in the mafia is not sufficient to predict social outcomes. That is, communities with similar initial social make-ups will not necessarily end up looking alike. This feature of the model sustains the claim that history does matter and that individuals newly entering the community may be pushed into replicating the choices of their predecessors in spite of their own preferences. Most important, this feature of the model is consistent with the observation that, depending on the nature and strength of a random shock to the system, a community may or may not develop a mafia. Second, the process is non-ergodic. This property is the key characterisic of path-dependency and is a necessary condition for the existence of social contagion. Randomly, a particular sequence of choices causes the process to bend toward a specific outcome among all the possible ones. A different sequence, however, would have bent it toward an alternative outcome. This feature of the model is consistent with the observation that mafia-like organizations become progressively more difficult to eradicate. Thus, nonpredictability describes individuals' interdependence by showing that their choices are influenced by what others choose. Simultaneously, no-ergodicity confirms the importance of each single individual contribution to the collective make-up of the community.
Policy implications
In a social process with no self-reinforcing properties-one with no social contagion-an exogenous change in the relative returns from alternative
OLD AND NEW MAFIA ORGANIZATIONS
43
activities is sufficient to affect future choices. With inceasing returns to adoption, however, this is not always the case. Indeed, once an activity starts to emerge as the dominant one, the possibility of choosing a different one becomes less and less attractive since the opportunity costs of that choice are inceasing. This means that if membership in the mafia begins to emerge as the dominant choice, the abatement of the organization becomes progressively more difficult to obtain. The adjustment necessary to shift the system back toward a relatively lower crime rate increases at an increasing rate and, beyond a critical threshold value, increases without bound. The possibility of the development and persistence of sub-optimal equilibria has important policy implications. Indeed, a pessimistic outlook emerges for the possible elimination of mafia-like organizations. From equation (9), the distance of the q(c) function from the diagonal is given by
D=~A
I(q(c)-c) ldc
where D may be interpreted as a measure of non-linearity. Thus, the higher D, the higher the degree of non-linearity. Non-linearity is a measure of the strength of the self-reinforcing mechanism existing in the process and, in this context, of social contagion. The more contagious the mafia is, the more difficult the implementation of successful anti-mafia policies becomes. This aspect of the model generates a testable hypothesis. Indeed, q(c) functions that have different Ds will generate time series data with different statistical properties. In particular, we can expect to find higher autocorrelation when D is higher, in other words when the self-reinforcing effect is stronger. Thus means that the correllelograph could be used to predict the difficulty of moving a community from a high-crime equilibrium with a mafia to a low-crime equilibrium without a mafia. By pointing out that the social costs of crime increase an an increasing rate, and the difficulty of eradicating well-established mafias, the model stresses the desirability of preventive measures, if only because they are cost efficient. l 2
Conclusion In Southern Italy, the mafia emerged from several powerful Sicilian families in the 1860s. In addition to the control and coordination of commerce and agriculture, these families enforced local order in areas where the newly established State was not able to enforce its laws. In Colombia, a mafia emerged in the 1970s to organize and control the shipment of marijuana and cocaine from Latin to North America. In Russia, the mafia
developed as a perverse outgrowth of the underground economy during the Soviet regime. In the case of Italy and Russia, the social system was shocked by a collapse of state control. In the case of Colombia, the system was shocked by a large-scale change in demand as foreign demand for illegal drugs rose sharply beginning in the late 1960s. The experience of different countries suggests that mafia-like organizations may originate from very different environments. Two common features, however, seem to exist. While the initial shock is country-specific and rooted in historical circumstances, the subsequent spread of the mafia presents characteristics which are similar across countries and may be described as a social contagion effect. First, a large-scale shock is necessary to bump the social system from a low-crime equilibrium without a mafia to a high-crime equilibrium with a mafia. Second, a network externality sets in pushing new individuals entering the economy toward choosing the mafia rather than honest work independently of their ex-ante preferences. This is so because membership in the mafia exhibits increasing relativereturns to adoption. To derive my results, I use a simple example of a general class of nonlinear path-dependent stochastic processes. In a dynamic context, my framework describes the choices of individuals who can produce income by either honest work or becoming members of the mafia. Given the existence of only two choices, I describe the returns to the two activities by using a family of relative-return functions all described by vertical displacement of the same function. The result is a single monotonically increasing function, q, mapping proportions into probabilities. My model shows that the larger the relative size of the mafia, the more additional people will choose to become members of it; that is, the mafia generates a self-reinforcing externality. Under certain conditions, the presence of social contagion may lock individuals into choosing crime independently from their natural preferences. In particular, the model identifies the existence of critical thresholds of crime incidence and shows how, and why, the crime rate tends toward specific equilibrium levels. If the mafia reaches the critical threshold size, the community tends toward, and ultimately settles at, a high level of crime. The model also suggests some policy implications. Such implications are different from those suggested by traditional crime models based on static frameworks. Suppose a community starts with a widespread mafia; for example, at C in Figure 2. Since point C is self-perpetuating, a policy aimed at reducing the size of the mafia would be successful only if able to make the system gravitate toward point A, the self-perpetuating mafia-free equilibrium. This means that if the policy fails to reduce the incidence of crime sufficiently, its effects will, at best, be transitory. Because attractive
OLD AND NEW MAFIA ORGANIZATIONS
45
equilibria correspond to stable fixed points o n the q function, which, in turn, is a linear transformation of the relative-return functions, the only way to push the process toward a mafia-free equilibrium requires a change of relative returns. But, because of the self-reinforcing property, to alter the relative-returns-to-membership in the mafia is difficult and costly. The idea hinges o n the non-ergodicity of the dynamic process, so that, after a certain number of individuals has chosen to become part of a mafia-like organization, the community is locked into a high-crime equilibrium and the adjustment necessary to shift it back toward a lower crime rate increases at an increasing rate.
Notes 1. Examples of such studies are Kindleberger (1983) on the narrow gauge of British railways and David (1985) on the Qwerty keyboard. 2. The percentage of economic activity controlled by the mafia represents an alternative measure. 3. "Competition between gangs is likely to reduce their number, which, at present, is thought to be 5,800 for the whole Russia," The Economist. July 1994, p. 20. 4. For a detailed mathematical treatment of non-linear path-dependent stochastic processes, see Arthur, Ermoliev and Kaniovsky (1983, 1987), and Hill, Lane and Sudderth (1980). 5. Returns are not only monetary, but include consideration of all costs and benefits associated with each income-producing activity. 6. For convenience I chose ("and rj"to be a third- and a second-order polynomial, respectively. Equations (1) and (2) are compatible with the intuition presented in the previous section and very easy to work with. The model and its results, however, are consistent with a wide class of functional specifications. Minniti (1995) contains a generalized version of this model, where the return functions are derived endogenously from the utility maximization problem solved by each individual. 7 . a, summarizes j's initial endowment; that is, the set of personal characteristics determining her relative-returns to membership in the mafia, independent from the relative size of the mafia in her community. Note that ai is equal to the difference between aj:'j's personal characteristics relevant to her returns from crime, and a;:' j's personal characteristics relevant to her returns from honest work. This is so because, although the two sets have many components in common, still some characteristics are relevant to the returns from one activity but not the others and vice versa. 8. Since there are only two choices, mafia or honest work, it would be redundant to make q a function of the two-dimensional vector showing the proportion of both criminals and non-criminals. 9. In non-linear path-dependent processes, a function q(x), describes the probability of each event type as a function of x , a vector recording the proportions of realized results of each type. Since q is a vector of probabilities and x is a vector
of proportions, q(x) maps a set of vectors onto itself; namely, the vectors whose elements sum to one and are all between zero and one. It is, of course, possible, for q(x) to have one or more fixed points. Arthur, Ermoliev and Kaniovsky (1983, 1987) proved that, under certain conditions, as the number of events increases, x tends to a limit vector x randomly selected from the set of all possible limit vectors. That is, they showed that shares of each event type converge with probability one to a stable fixed point of q(x. 10. The following paragraph is a version, modified to fit my model, of the general explanation of the dynamics of path-dependent processes given in Arthur, Ermoliev and Kaniovski (1987). 11. These properties are extensions of the probabilistic characteristics of all nonlinear path-dependent processes. 12. This property strongly supports the conclusions drawn by Crane (1991) in his epidemic theory of ghettos. Although Crane's analysis refers to high-school dropout rates and teenage pregnancy, the type of dynamics he describes is very close to that of mafia-like organizations.
References Arthur, W. Brian. 1989. "Competing Technologies, Increasing Returns, and Lock-in by Historical Events," The Economic Journal, vol. 99, pp. 116-31. -, Y. M. Ermoliev and Y. M. Kaniovski. 1983. "On Generalized Urn Schemes of the Polya Kind," Cybernetics, vol. 19, pp. 61-71. ----, Y. M. Ermoliev and Y. M. Kaniovski. 1987. "Path-Dependent Processes and the Emergence of Macro-Structure," European Journal of Operational Research, vol. 30, pp. 294-303. Becker, Gary S., and William M. Landes (eds.). 1974. Essays in the Economics of Crime and Punishment. New York: Columbia University Press. Brower, George D., and Isaac Ehrlich. 1987. "On the Issue of Causality in the Economic Model of Crime and Law Enforcement: Some Theoretical Considerations and Empirical Evidence," The American Economic Review, vol. 77, no. 2, pp. 99-106. Case, A,, and L. Katz. 1991. "The Company You Keep: The Effects of Family and Neighborhood on Disadvantaged Youth," NBER Working Paper 3705. Crane, Jonathan. 1991. "The Epidemic Theory of Ghettos and Neighborhood Effects on Dropping Out and Teenage Childbearing," American Journal of Sociology, vol. 96, no. 5, pp. 1226-59. David, Paul. 1985. "Clio and the Economics of the QWERTY," American Economic Review Proceedings, vol. 75, pp. 332-7. Gambetta, Diego. 1992. La Mafia Siciliana. Einaudi, Torino. Granovetter, M., and R. Soong. 1983. "Threshold Models of Diffusion and Collective Behavior," Journal of Mathematical Sociology, vol. 9, pp. 165-79. Hill, Bruce M., David Lane, and William Sudderth. 1980. "A Strong Law for Some Generalized Urn Processes," The Annals of Probability, vol. 8, no. 2, pp. 214-26.
OLD AND NEW MAFIA ORGANIZATIONS
47
Jennings, William P. 1984. "The Economics of Organized Crime," Eastern Economic Journal, vol. 10, no. 3, pp. 315-21. Kindleberger, C. 1983. "Standards As Public, Collective and Private Goods," Kyklos, vol. 36, pp. 377-96. Minniti, Maria. 1995. "The Dynamics of Social Contagion: The Case of Crime," mimeo. New York University. Miyao, T. 1978. "Dynamic Instability of a Mixed City in the Presence of Neighborhood Externalities," American Economic Review, vol. 68, pp. 454-63. Pyle, David J. 1983. The Economics of Crime and Law Enforcement. New York: St. Martin's. Sah, Raaj K. 1991. "Social Osmosis and Patterns of Crime," Journal of Political Economy, vol. 99, pp. 1272-95. Schelling, Thomas C. 1967. "Economic Analysis and Organized Crime," in the President's Commission of Law Enforcement and Criminal Justice, Task Force Report: Organized Crime. Washington, D.C.: U.S. Government Printing Office. The Economist. "Russia's Mafia: More Crime than Punishment," July 9, 1994, pp. 19-22.