University of Southern California Dissertations and Theses

An economic analysis of cooperative group behavior: Theory and evidence

(USC Thesis Other)

An economic analysis of cooperative group behavior: Theory and evidence

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content AN ECONOMIC ANALYSIS OF COOPERATIVE GROUP BEHAVIOR: THEORY AND EVIDENCE by Mussaddequddin Chowdhury A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (Economics) January 1993 Copyright 199 3 Mussaddequddin Chowdhury UMI Number: DP23396 All rights reserved INFORMATION TO ALL USERS The quality of this reproduction is dependent upon the quality of the copy submitted. In the unlikely event that the author did not send a complete manuscript and there are missing pages, these will be noted. Also, if material had to be removed, a note will indicate the deletion. Dissertation Publishing UMI DP23396 Published by ProQuest LLC (2014). Copyright in the Dissertation held by the Author. Microform Edition © ProQuest LLC. All rights reserved. This work is protected against unauthorized copying under Title 17, United States Code ProQuest LLC. 789 East Eisenhower Parkway P.O. Box 1346 Ann Arbor, Ml 48106-1346 UNIVERSITY OF SOUTHERN CALIFORNIA THE GRADUATE SCHOOL UNIVERSITY PARK LOS ANGELES, CALIFORNIA 90007 This dissertation, written by M(JSsA7>vi=:$uT)vt/J £ HoifjQtHOf?y under the direction of h./.A.. Dissertation Committee, and approved by all its members, has been presented to and accepted by The Graduate School, in partial fulfillm ent of re quirements fo r the degree of 713 D O C T O R O F P H ILO S O P H Y Dean of Graduate Studies Date .... DISSERTATION COMMITTEE rson DEDICATION To the loving memory of my father Moinuddin Chowdhury and with love and gratitude to my uncle Abdus Satter Chowdhury ACKNOWLEDGEMENTS j The many people who have helped, directly or | indirectly, in the process leading to the writing of I this dissertation are too numerous to be enumerated 1 completely. I read somewhere a long time ago about the 1 ideal Greek concept of a teacher as friend, philosopher, and guide. I did not realize the exact ' significance of this idea until I met and worked under my dissertation committee chair, Professor Jeffrey B. | Nugent. Without his active intellectual stimulation as well as concern for my financial needs, this dissertation would not have been possible. I have no count of the numerous drafts he read of this work and the prompt and valuable comments and words of criticism and encouragement he offered me over the years. I find myself at a loss for words to express my gratitude to him. Suffice it to say that no graduate student could have hoped for a better advisor: kind, understanding, and yet uncompromising with respect to the quality of work. The other members of my committee have also contributed to the successful completion of this work. iv From Professor Richard A. Easterlin, I first recognized the importance of consideration of sociological and political factors in addition to economic ones in the process of economic development. Having had the privilege of working as a teaching assistant for him for one semester, I had the opportunity of observing an excellent teacher at work. Professor Timur Kuran's courses and articles strengthened my conviction of the superiority of the interdisciplinary approach to social science. I am grateful to Professor Maurice D. Van Arsdol and the Population Lab for a summer grant and for the Hugo R. Santora Endowed Fellowship while this dissertation was being written. I also wish to express my debt to Professor Judith Treas, a former member of my committee, who left for University of California, Irvine. Finally, it would not have been possible for me to pursue my graduate studies here without the financial support in the form of fellowships and teaching assistantships from the economics department and the Graduate School and a fellowship from the Haynes Foundation. Turning now to the people outside the committee, I must start with my family in Bangladesh: my brothers, V i sisters and cousins who encouraged me every step of my I way. My mother, Amirunnessa Begum, would have been i i very happy had she lived about 27 years longer. My ! aunt, Sofia Sultana, who died in 1991, is sorely missed i | and gratefully remembered. Of the many wonderful i | people I met in Los Angeles as fellow graduate ; students, friends, departmental staff and USC staff, I particular mention must be made of the following: Mohamed Bechri, Tamim Ahmed, Tom Hollihan, Patti Riley, t i Gazala Pirzada, Mario Galevski-Dimitrovsky, Hedi | Trabelsi, Bibhas Saha, Weihong Huang, Ghaffar Mughal, ■ Elizabeth Zelinski, Mike Gilewski, Subroto Sarkar, ; Vivian Smith, and Carol Robertson. My wife's sister ! Daisy Shamim and her husband C. M. Shamim were the only I relatives we had m the Los Angeles area and their companionship relieved some of the pains of homesickness. Apart from writing this dissertation, the best part of my graduate student days is the many friends we made, several of whom will remain life-long friends. During the final year of the writing of this dissertation, I accepted a position at Humboldt State i i University. My colleagues in the economics department, John Grobey and Jacqueline Kasun, were extremely supportive of my efforts to complete my degree. The computer consultant of the School of Professional Studies, Joan Van Duzer, facilitated the transition from the system at USC to that at HSU. Finally, I acknowledge with gratitude the hospitality of Patti Riley and Tom Hollihan and the Shamims during the long periods of time my family and I were house guests during the summer of 1992 and winter of 1992-93 when this work was in the final stage of completion. Without the love and friendship that these two families displayed to us in numerous tangible and intangible ways, life in Los Angeles would have been infinitely more difficult. The two people who shared with me the trails and travails of graduate study were my wife Najat and my daughter Najwa. The long hours of working at the library and/or at the computer center inevitably meant the physical absence of a husband or father. Even when I was physically present but was trying to check the consistency of a model or analyze the results of an empirical test mentally, questions asked or conversations initiated visually went unanswered. Najwa vii was less than four years old when I started my graduate work; by being extremely quiet and well-behaved she made it possible for me to work in our small one- bedroom apartment. Finally, I turn to two brothers who taught me from my first faltering footsteps the prime importance of seeking knowledge. My deceased father, Moinuddin Chowdhury, encouraged me to buy and read as many books as possible and instilled in me the conviction that nothing is more worthwhile to do in our lives than the relentless pursuit of knowledge. Recognizing the limitations of the availability of higher education in Bangladesh, he always advised me to go abroad. After his death in 19 66, my uncle Abdus Satter Chowdhury continued where his elder brother had left off. Family obligations and priorities necessitated my staying in Bangladesh after completion of studies there. But, at every opportunity, my uncle did not hesitate to let it be known that he would like for me to pursue graduate studies abroad. To a large extent my decision to pursue a doctoral degree was influenced by these two remarkable men. It has been my extremely good fortune to have had two people such as these as father and Vlll uncle and it is to them that I dedicate this dissertation. Mussaddequddin Chowdhury Los Angeles, California. TABLE OF CONTENTS Chapter Title Pages 1 Introduction 1-25 2 General Review of the Literature 26-92 3 A Formal Model of Group Behavior 93-156 4 The Grameen Bank in Bangladesh: 157-203 Why Groups Succeed 5 Norms and Mobility 204-264 6 Conclusion 265-310 ABSTRACT This dissertation consists of six chapters and i examines the theory and evidence of cooperative group i j behavior. The introductory chapter distinguishes j between methodological individualism and methodological j holism and chooses the former as the method of J analysis. Cooperation is achieved by rational, self- ! interested individuals and altruism is shown to be I neither necessary nor sufficient for group success. The second chapter provides a representative, though selective, review of the literature on group behavior from the fields of economics, sociology, political science, and evolutionary biology. In ’ addition to the seminal work of Olson, those of Hardin, i I Coleman, Hechter, Heckathorn, Axelrod, Granovetter, i Basu and Maynard Smith are reviewed. i The third chapter presents a recursive model to I help understand the effect of social control on human behavior. These controls are mathematically similar to other types of constraints, such as budget constraints. They are modelled as predetermined variables that undergo changes at (almost) every iteration, thus | providing the framework for the study of norm formation i I and decay. | Empirical evidence is provided in the fourth ! chapter to support the contention that formal credit j schemes are failures in most developing countries. The | Grameen Bank in Bangladesh is a notable exception and I it is demonstrated that this phenomenal success can be I attributed to the design of an institution that capitalizes on the existent norms of the society. i j In the fifth chapter, an Indian dataset is used to I | make an econometric analysis of the impact of mobility j on a social norm like old-age support of parent(s) by children. The results confirm the model hypothesis that mobility reduces social control which is measured as the likelihood of caring for elderly parents. They also show that mobility has smaller and less certain I effects on caring for the mother than it does for the father. i The concluding chapter shows how a small i difference in norms and beliefs may lead to a large difference in group behavior. Thus, replicability of a ; successful institution from one country to another must be carried out with proper and careful attention to such differences. Finally, emphasis is placed on the need for multidisciplinary study of human behavior involving economists, sociologists and anthropologists. CHAPTER 1 INTRODUCTION 1.1 Background The purpose of this dissertation is to examine the economic bases for cooperative group behavior and to see whether empirical evidence supports the theoretical results. Among others, sociologists and economists have tried to study group behavior. The methodology used here is choice-theoretic or "economic", in the sense that explanations are grounded in individual choice. Each individual is a rational, utility-maximizing agent and group behavior is an outcome of the individuals' choices. While psychology and perhaps other branches of knowledge also consider individual choice, the difference between them and the economic one taken here is primarily one of emphasis. While psychology and certain other fields study why certain norms evolve, economics is more concerned with the ramifications of the norms. Reference will be made to work done by scholars in psychology, sociology and evolutionary biology; a systemmatic study of all 2 branches of knowledge, however, is beyond the scope of this dissertation.1 1.2 Methodological Individualism Versus Methodological Holism As Coleman [1988] points out, "[there] are two broad intellectual streams in the description and explanation of social action." One— which may be called the sociological view— sees the actor as completely socialized and his actions determined by social norms, rules and obligations. Its comparative advantage is in attaching importance (undue importance, in my view) to the social constraints individuals typically face. The problems are at least two-fold: i) individual agents are powerless and are in effect impotent robots and, thus, ii) there need be no connection between individual action and group action except in the sense that the latter determines the former. An even more fundamental criticism is that such a theory may be functionalist: any behavior at the 1The primary focus will be on the study of group behavior by the social sciences like economics and sociology. 3 individual level may be attributed to some hypothetical social control without any attempt at explaining either the existence of the control or the individual's unquestioning acquiescence to such control. Not all sociologists, however, subscribe to what I call the sociological view. One of the earliest examples of this stream was Homans [1958]. The title of one of his papers, "Bringing Men Back In" is self-explanatory. Among modern-day sociologists, Coleman, Hechter, Heckathorn and Granovetter exemplify scholars who take the purposive action of individuals as their starting point and then examine how group (or social) behavior ensues, and how individuals are affected by social constraints.2 According to the above classification used here, these sociologists belong to the "economic" school. The other intellectual stream— "economic"— pictures a rational actor having goals (or preferences) and acting independently and in a self-interested 2The works of these scholars are discussed in Chapter 2. 4 manner to maximize his satisfaction (or utility).3 i The major criticism of this view is that the social context is completely ignored. It is patently unrealistic to assume that norms, interpersonal trust and social institutions play no role in affecting individual behavior. But the economic view need not consider the individual as devoid of social identity. Much of the criticism of the economic analysis is based on a misunderstanding of the concept of rationality. The textbook neoclassical economic model pictures an egoistic, rational individual. As Brennan and Pincus [1987] point out, the reason agents are assumed to be egoistic is to enable us to examine the extent to which the institutional framework (the incentive structure, for example) can convert 3Note that the "economic" and "sociological" methods respectively to methodological individualism and methodological holism. The terms are used to denote two ways social science views social behavior. For example, it should not be construed to mean that all (or even most) sociologists adopt methodological holism at the present time but that that is the bias of the discipline. In fact, as discussed in Chapter 2, many sociologists have the individual as their unit of study and have thus adopted methodological individualism. 5 selfishly4 motivated actions to those that benefit everyone. Thus, Adam Smith's remark that we do not depend on the benevolence of the brewer and the baker for our beer and bread but on their own selfish interests does not mean that the brewer or the baker cannot be altruistic. (S)he is interested only in the market incentive structure that renders selfish actions socially desirable. 1.3 Rationality The assumption of rationality concerns the structure of preferences, e.g., that the preference ordering is complete and transitive accompanied by a (constrained) maximizing principle, i.e., every agent is doing the best he can. Rationality presumes nothing about the content of preferences.5 Utility functions 4It is probably preferable to denote such behavior "self-interested" rather than "selfish". The former could include altruistic behavior not because it helps others but because it increases the utility of the rational individual. 5Frank [1990] criticizes what he perceives as the shortcomings of the rational choice model. He thinks that not only do people make decisions based on incomplete information (which, incidentally, is rational if the costs of gathering information is high) but also that people make poor use of the information they have. Although he provides several examples of 6 can contain anything as arguments. An egoistic person's utility function would be: (1) Uf = U^X.) , U‘ > 0 where X , - is the bundle of goods he consumes. On the other hand, if i cares about j, i's utility function would be: (2) Uf = Uf(X,, Xj), Uxi > 0, UxJ > 0. An extreme altruist would have a utility function like: (3) U, = U i(Xj), Uj > 0. Rationality does not specify what type of utility function a person has.6 It merely postulates that, given a certain type of utility function, an agent will what he terms "irrational behavior with regret", it is not clear to me that these types of behavior are widespread, persistent and systemmatic. Poor processing of information may, for example, lead one to do more of some activity and another to do less and the "average" prediction based on rational choice theory may still be valid. The second objection of Frank's is what he calls "irrational behavior with regret" where people people behave not only out of reason but also out of passion. We do not regret having left a tip in a roadside restaurant far from home even though the waiter cannot retaliate if we did not leave the tip. What Frank seems to be saying is the necessity of broadening our utility function to include nonmonetary payoffs and, in the model in Chapter 3, we do precisely that. Note that rationality does not imply that only monetary payoffs need be considered; it merely states that individuals do the best they can. 6Each of the above types of preferences may be termed "self-interested". 7 attempt to maximize in a consistent way, subject to whatever constraints one faces. The typical constraint economic theory considers is the budget constraint. There is, however, no methodological problem in handling other constraints. Suppose there is a law against smoking in public places. Then the choice set is restricted; in particular, the good (or bad) "public smoking" is no longer available. Preferences defined over this good are not relevant in terms of choice. Given this change, the economic analysis can continue as before. It is precisely adjustments like this that need to be made in order to consider group (or social) action. As will be demonstrated in a formal model, the existence of (social) controls restricts or enlarges the choice set. It needs to be emphasized that, unlike the results obtained from textbook economic analysis, controls do not always restrict the choice set. In many cases, as discussed in Chapter 2, norms or social control may constitute what Coleman [1988] calls "social capital" that facilitates economic activity. It is most convenient to view such norms as devices through which 8 information, monitoring and sanctioning costs are lowered. In this connection, the work of Easterlin [1973] is important. He demonstrates that a rise in a particular individual's real income typically brings more happiness to that individual but that a rise in everyone1s income does not increase everyone1s happiness. Individuals evaluate how well off they are not in terms of the absolute amount of goods and services they enjoy but relative to "... a social norm of what goods they ought to have." While economic considerations are most often cited as contributing to happiness, so are family considerations, health, and social and political conditions.7 Easterlin is an example of an economist who stresses the importance of social norms in behavior but assumes a utility-maximizing, rational individual. Basu [1984] contains a concluding chapter entitled "The 7In such cases, the utility function may be represented as: (4) Uf = Ujfx,), with U' > 0, where x , - = X5 /X with X- is the individual's consumption and X is one's norm of what one should have. Thus, one's utility goes up only when one's consumption relative to one's expectations goes up. Limits of Economic Analysis". He rejects the view that social institutions are mere decorations and accepts Akerlof's8 theory that institutions like the caste system are exogenous and may persist because it is in the interest of individuals to abide by the rules. The result would be a low-level equilibrium trap. Basu also notes that norms and custom may impede mobility and innovation. This work is an example of the typical (and incorrect) view most economists display when they view all norms and customs as necessarily restricting the choice set and is not in the same vein as Easterlin's work. Note that Basu [1991] has made the transition from viewing norms as a constraint to human behavior to viewing them as satisfying what he calls "minimal functionalism" which is the opinion that a norm that persists cannot be too harmful to society.9 Kuran [1987] suggests, however, that social institutions (like the caste system) can persist and ®Akerlof, G. 1976. "The Economics of Caste and of the Rat Race and other Woeful Tales". Quarterly Journal of Economics, 90. 9Basu [1991] is discussed in Chapter 2. It seems that Basu position over the years has moved closer to that of Easterlin. 10 even be publicly supported by its apparent victims because the costs of publicly opposing them were initially high and, over time, propaganda was used to convince people to believe that such an institution is beneficial. Thus, an institution which is harmful may well persist. Kuran [1989] demonstrates that unpopular institutions may, however, be very fragile. He shows, using the example of the Shah of Iran, how an unpopular government which appears formidable can crumble with a slight increase in the opposition's apparent size.10 He takes a critical mass approach where people will not publicly oppose an institution (even if they dislike it) until a certain proportion or number of other people1 1 also do so. Under these conditions, the outcomes can be different for very similar groups. Granovetter [1990] observes, "... outcomes can vary dramatically ... if the social structure, institutional 10The recent experiences in the former Soviet Union and in Eastern Europe provide further evidence of this. 11hence the term "critical mass" 11 history, and collective action are different.12 As shown in Chapter 6, a slight difference between the cultural norms of Malaysia and Bangladesh (which have many similarities) led initially to the failure of the attempt to replicate the highly successful Grameen Bank (GB) in Malaysia. 1.4 Prisoners' Dilemma The starting point for studying group behavior is the Prisoners' Dilemma (PD). In its pure form, the payoff matrix of a PD situation may be depicted as: 12Consider a situation where utilities are interdependent such that there are n people in a group and the threshold of each individual i is t- = i - 1. More specifically, a person i will demostrate in the street if (i - 1) people do so. Thus, the first will take to the streets even if no one else does (1 - 1 = 0), the second will do so as soon as the first one does, and so on. The result is a big street protest with all n people taking part. Consider a slightly different preference ordering where : t. = i - 1 for i = 1, and 3, 4, ..., n and t? = i for 2 The first person will be out in the street even if no one else is. But the second person will not do so until there are two people. In this group, therefore, there will be only one person demonstrating in the street. Note, however, that the two groups are almost identical in their disposition. Note, further, that that a change in the disposition of just one person (person 2) can dramatically can the outcome. 12 Person B Cooperate Defect R = 3 S = -2 Cooperate R = 3 T = 5 Person A T = 5 P = -1 Defect S = -2 P = -1 Figure 1.1 If both players cooperate, they get R, the reward for mutual cooperation. If one player cooperates while the other defects, the cooperator gets S, the sucker's payoff, and the defector gets T, the temptation to defect. Finally, if both defect, they each get P, the punishment for mutual defection. The best a player can do is T and the worst is S. R is assumed to be better than P. Thus, we have: (i) T > R > P > S. It is further assumed that an even chance of exploiting and getting exploited is not preferable to mutual cooperation: (ii) R > ( T + S )/2. Conditions (i) and (ii) together define a pure PD situation. Thus, we have a situation where individual 13 egoistic behavior leads to a socially desirable result. Two egoists playing this game once will choose their dominant strategy, defection, and each will get less than (s)he could have obtained had both cooperated. Adam Smith's invisible hand translating individual welfare into social welfare does not seem to work. What will happen if the game is played many times? If it is played a known, finite number of times, there is still no incentive to cooperate. But, as Axelrod [1980a and 1980b] has demonstrated, cooperation can occur with an indefinite, finite number of iterations provided the agents care enough about the future. My own interests in this scenario are a little different. It is the first round of interaction which is crucial and largely overlooked in the literature. Axelrod [1980a and 1980b] and Rapoport [1983], among others, have demonstrated that cooperation can emerge if the game is played an unknown number of times. But, why should a player cooperate with another (unknown?) player on the first round? Once cooperation is attained in the first round, it appears that a major obstacle to group failure is removed. As soon as tangible gains are obtained (and they will be if 14 cooperation occurs in a PD situation), it certainly appears more likely that cooperation will result in the future rounds. 1.5 Norms and Cooperation To help explain cooperation on the first round, institutional factors must be examined. It seems that social control— external or internal— is a major factor that facilitates (but does not assure) successful group behavior. An external form of social control would involve pecuniary or non-pecuniary sanctions against defection from some social norm or custom. An internal form of social control would involve internalization of the norm and would cause psychic distress from its non-observance. An example would be the situation of a Muslim drinking alcohol. If (s)he does not care for the religious injunction against drinking alcohol but does not do so because (s)he is living in a fundamentalist Islamic country (even though (s)he enjoys cognac), cooperation (in not drinking alcohol) is attained through external sanctions. On the other hand, if a Muslim does not drink even in a foreign country (because (s)he believes it is not the right 15 thing to do), the norm has been internalized. As is obvious, cooperation is more easily obtained and is less costly where norms are internalized. Casual empiricism tells us that these kinds of norms are commonplace. There are orthodox Jews and Muslims who do not eat pork. There are Hindus who believe that the caste system is the natural order of the universe. Some of these norms are backed by sanctions. In many societies, there are taboos against taking interest or using contraceptives. Moreover, sometimes these norms evolve into laws, e.g., monogamy and usury laws. It is not the purpose of this dissertation to examine how and why these norms evolve but merely to acknowledge that they exist and operate as constraints on human behavior. The word ••constraint" should be interpreted in a neutral sense. Some norms (e.g., monogamy, anti-trust laws) may be beneficial to society while others (e.g., worship of elders, the caste system) may be harmful, at least under some circumstances.13 Standard economic analysis either abstracts from the constraints that norms impose or refers to them almost as an afterthought. The sociological view attaches pre-eminent importance of adherence to such norms as the "normal" behavior. What this dissertation demonstrates is why certain groups, who apparently have no (economic) reason to cooperate, do so. The assumption is that of a rational individual who maximizes utility while living in a certain society and thus a person who is affected (though not necessarily controlled) by social norms and customs. If the controls are of the type that increase economic and/or psychic benefits, they would tend to be perpretrated. If, on the other hand, controls reduce economic and/or psychic payoffs, they would tend to be eroded. This will be explained by the model in Chapter 3 which will provide a framework for examining not only the impact 13The textbook economic analysis almost always views norms as harmful to society (see above footnote). Thus, it would be unable to explain why a norm like "pay off all debts" could in fact enlarge the choice set by making possible mutually beneficial intertemporal trade (see Chapter 4). 17 of norms on human behavior but also how norms may come into existence or erode over time. Yet, as will be explained in Chapter 3, the model is probabilistic and thus does not degenerate into functionalism. It is merely more likely that a beneficial norm will persist, although there may well be many harmful norms which do not wither away. Let us examine some cases of cooperative group behavior which appear puzzling: (a) Why do people vote? (b) Why can millions of dollars of transactions occur in wholesale diamond markets without any formal record-keeping? (c) Why do people in Mexico City think nothing about throwing candy wrappers in the streets while such behavior would be unthinkable in Tokyo? (d) Why does a member of a rotating-credit association in India continue to contribute even after he receives an early payout? (e) Why does a man think it is right (even obligatory) for him to censure his neighbor's son in rural Bangladesh while such behavior may be wrong (even undesirable) in Los Angeles? 18 (f) Why does the Grameen Bank in Bangladesh— which lends only to landless villagers— have a repayment rate that Chase Manhattan would be proud of? Why don't other banks in Bangladesh have even remotely comparable repayment rates? (g) Why do people in traditional societies look after their parents in the latters' old age? 1.6 Altruists' Dilemma An appeal to altruism might appear to be able to solve many of these puzzles. But, whether altruistic behavior pertains in a particular setting is an empirical matter and difficult to test. It seems safe to assume that selfish behavior is more common than altruism. Moreover, it should also be noted that altruism is neither a necessary nor a sufficient condition for successful group behavior. The following payoff matrix which (Ullmann-Margalit, 1977) calls Altruists' Dilemma illustrates this: 19 Person B Sit Don't Sit Sit 0 0 1 0 Don't -2 -1 Sit 1 -1 Figure 1.2 There is one unoccupied seat on a bus which can (uncomfortably) seat both persons. Both, however, refrain from sitting there with the altruistic intention of leaving it free for the other. The dominant choice is thus ( -1, -1 ). Altruism is not sufficient for attaining the social optimum. As is obvious, it would be sufficient if only one of the two people were altruistic. But, in this Altruists' Dilemma, there is too much altruism. Altruism may also not even be necessary for social optimum because, as will be shown in Chapter 3, even when they behave in a purely self-interested manner, rational individuals can achieve cooperation. What this requires is merely a longer planning horizon where short-run costs are incurred in anticipation of long- run benefits. Thus, one contributes to group success 20 not because one cares about the welfare of other members but because one anticipates future benefits for oneself. Furthermore, as discussed in Chapter 5, there is what I term the "disguised altruist" whereby an apparently altruist behavior is, in reality, based on selfish motives. People may work in government at reduced pay not because they want to sacrifice for the sake of societal welfare but because they enjoy proximity to the levers of power. It should also be noted that some features of altruism may be built into laws, e.g., tax deductibility for charitable contributions. 1.7 The View Ahead This dissertation has six chapters. Chapter 2 contains a review of the literature. The seminal work is that of Olson. Extensions of Olson's work in Hardin will be examined. Contributions by Coleman, Hechter, Heckathorn, Granovetter, Axelrod, Maynard Smith, and Basu will also be discussed. Specific citations of these works will be found at the end of Chapter 2. These works run the gamut from Olson's pessimism about group success to the optimism of Maynard Smith. 21 Coleman shows how social controls can contribute to group success and enlarge the choice set. Chapter 3 presents a formal model incorporating social control in economic behavior. From this model, comparative statics results are derived. The model incorporates Heckathorn's notion of social control with what Coleman calls social capital. More importantly— and this is the main contribution of the model— it takes into account the individual preferences that affect human behavior. Social controls are exogeneous to the model but affect individuals' propensity to cooperate in a group situation. Chapter 4 studies the causes of the success of the Grameen Bank (GB) in Bangladesh. Empirical evidence is presented which supports the contention that formal credit schemes are failures in most less-developed countries (LDCs). Repayment rates are almost universally low and credit seems to flow disproportionately to the richer farmers. Exceptions to this general pattern of failure are the successes of group credit schemes like the GB and the rotating-credit associations. It will be argued that 22 social institutions are responsible for this difference. Chapter 5 treats the impact of mobility on social norms like the provision of old-age support by children in a traditional society like India. In such societies, the lack of alternative marketable savings causes people to look to their children for support in their old age.14 The model developed in Chapter 3 predicts that the viability of these conventional forms of social security hinges (rather precariously) on the existing norms and on the children's expected benefits in the future. If increased mobility leads to reduced social control, economic modernization (which is typically positively associated with mobility) should erode social controls and thus lead to a situation where fewer people will care for their old parents. Furthermore, economic incentives tend to lower the demand for children. We can visualize a situation where, as the economy grows, there are fewer children and they are less likely to care for their elderly parents. Thus, ceteris paribus. this informal 14See, for example, Nugent [1985]. 23 institution of old-age support is likely to erode over time. Chapter 6 examines the policy and other implications. There are both static and dynamic ramifications. It demonstrates how a small difference in norms and beliefs may lead to a big difference in the success/failure of an institution. In the case of GB, some forms of social control (e.g., social ostracism) may be performing a useful function. But, as society becomes more mobile, such norms are likely to be less effective. In the case of old-age security, the urgent need for formal social security schemes is obvious. In the interim period, it is probably advisable to devise incentive schemes to maintain the traditional mechanisms of support. 24 BIBLIOGRAPHY Axelrod, Robert. 1980a. "Effective Choice in the Prisoner's Dilemma.1 1 Journal of Conflict Resolution 24: 3-25. ---------------. 1980b. "More Effective Choice in the Prisoner's Dilemma." Journal of Conflict Resolution 24: 379-403. Basu, Kaushik. 1984. The Less Developed Economy. Basil Blackwell, New York. --------- . 1991. "Civil Norms and Evolution". unpubished paper. Brennan, Geoffrey and Jonathan Pincus. 1987. "Rational Actor Theory in Politics: A Critical Review of John Quiggin." The Economic Record, Volume 63, No. 180: 22-32. Coleman, James S. 1988. "Social Capital in the Creation of Human Capital." American Journal of Sociology 94 Supplement: S95-120. Easterlin, Richard A. 1973. "Does money buy happiness?". The Public Interest, Number 30, Winter 1973. Frank, Robert H. 1990. "Rethinking Rational Choice", in Roger Friedland and A. F. Robertson (eds.), Bevond the Marketplace. Aldine de Gruyter, New York, pp. 53-87. Granovetter, Mark. 1990. "The New and Old Economic Sociology", in Roger Friedland and A. F. Robertson (eds.), Bevond the Marketplace. Aldine de Gruyter, New York, pp. 89-112. Homans, George. 1958. "Social Behavior as Exchange." American Journal of Sociology 63: 597-606. Kuran, Timur. 1987. "Preference Falsification, Policy Continuity and Collective Conservatism", The Economic Journal, 97: 642-665. 25 -------------- # 1989. "Sparks and Prairie Fires: A Theory of Unanticipated Political Revolution", Public Choice, 61: 41-74. Nugent, Jeffrey B. 1985. "The Old Age Security Motive for Fertility." Population and Development Review, 11: 75-97. Rapoport, A. and A. M. Chaitimah. 1965. Prisoner's Dilemma. University of Michigan Press, Ann Arbor. Rapoport, Anatol. 1983. Mathematical Models in the Social and Behavioral Sciences. Wiley, New York. Ullmann-Margalit, Edna. 1977. The Emergence of Norms. Oxford University Press, Oxford. CHAPTER 2 GENERAL REVIEW OF THE LITERATURE 2.1 Group Cooperation and Individual Maximization It was taken for granted by most economists that group welfare would be maximized by each individual's self-interested maximizing behavior. The tradition follows from Adam Smith's "invisible hand". Two or more individuals— with different utility functions and/or initial endowments— could always achieve a Pareto-superior position by voluntary exchange. And, after exchange, the position attained is Pareto optimum, that is, no one's utility can be increased without decreasing that of another. Basically, all this is a logical extension of the theory of comparative advantage. Considerable attention was then devoted to the construction of a social welfare function. Changing the endowments of individuals led to a different Pareto optimum and, in the absence of such a function, it was not possible to evaluate the two situations. Such details do not concern us here. 27 What is relevant for our purpose is the customary depiction of individuals and firms, especially the former. All types of agents are assumed to be available in large numbers, making each a price taker. This assumption, together with others that imply perfect competition, is unobjectionable in that the results would serve at worst as a benchmark. But the problem is that the individuals are also assumed to be devoid of any social context: they possess no culture, no history, no beliefs. This is the main criticism of other social scientists, especially sociologists. As will be demonstrated in Chapter 3, explicit consideration of the social context can be undertaken within the framework of methodological individualism. One does not necessarily have to appeal to norms or altruism to explain successful group behavior, although in certain contexts these may be helpful or even sufficient for such success. A common problem encountered in writings about aggregate behavior was the fallacy of composition. It was thus thought that, if individuals maximize utility, so, too, would groups of individuals. But, as Olson 28 [1965]1 showed, a public good which would increase the utility of a group is not likely to be provided at all by individuals maximizing utility. Similarly, in Marxian theory, it was assumed that the revolutionary working class would further its (corporate) interests.2 We thus encounter two separate issues: (i) whether group cooperation will necessarily occur just because members of the group are benefited by a particular course of action and (ii) whether individuals are truly "atomized", that is, devoid of any social context. Most economists— especially following the publication of Olson [1965]— agree that the answer to (i) is "no". But they are conspicuously silent about (ii). Sociologists, on the other hand, are usually silent about (i) but answer (ii) with an emphatic "no". On a continuum of "atomized" to "socialized", sociologists, 1See below for details. 2In these cases, it is essentially the free rider problem which precludes the provision of public goods (see below for details). See also Easterlin [1973], discussed in Chapter 1, however, where raising everyone's real income does not increase everyone's happiness because of one's expectation of what one should get. 29 in general, assert that individuals are completely socialized. Thus, group cooperation can be attributed to norms, conventions, and beliefs. Since such a theory cannot explain non-cooperation, it denotes such behavior as "deviant". The individual is viewed as merely a puppet in the hands of the puppeteer, the society. Sociological literature which emphasizes the completely socialized view of individuals is not reviewed because it is not directly relevant to the model developed here but merely to the phenomena the model seeks to explain. Furthermore, the essential arguments of this literature have already been stated in the previous paragraph. In this context, it is also important to note that only sociologists of the methodological individualism school have considered (ii) explicitly in an interesting manner. This does not mean that some economists have not recognized the need for multidisciplinary studies. One of the earliest and strongest proponent of this need has been Easterlin whose works over the years have emphasized this need. Easterlin [1973] showed that, while economic considerations predominate, family 30 considerations, health, and social and political conditions all contribute to human happiness. Easterlin [1978] criticizes "the undue narrowing of "economics of fertility", and a consequent alienation from allied work by sociologists and biologists." He then attempts to make the economic analysis more compatible with biologists and sociologists. Easterlin and Crimmins [1991] identified four types of life goals: (a) private materialism or one's own material well-being, (b) family life or goals relating to family formation and welfare, (c) public interest or concerns about the broader community, and (d) personal self- fulfillment. Note that standard economic utility maximization problems are limited, in the main, to (a) and that (a) and (d) would be considered synonymous.3 In our frame of things, (c) may be thought of as the social context. 3But, see Easterlin [1973] above and in Chapter 1. 31 2.2 The Main Contributors A general survey of the literature is presented below, drawn from economists, political scientists, sociologists, and biologists. The survey is highly selective and thus not comprehensive but it is certainly representative. The main contributors from each field are chosen and analyzed and their interlinkages pointed out. 2.2.1 OLSON Olson [1965] is the pioneering work with respect to (i) above. He demonstrates that, unless the number of individuals in a group is small, or unless there is coercion or selective incentives, "rational self-interested individuals will not act to achieve their common or group interests." This result holds even when there is no information problem, that is, when it is clearly and universally known to the individual members that a particular collective action will benefit each member, and there is no disagreement about the methods of achieving it. The logic of this result stems from the free-rider problem which arises when each of the parties to a mutually agreeable outcome has an incentive to walk out of the agreement. 32 Everyone desires police protection and would agree to the need of a police force. An individual would not, however, hesitate, to evade the taxes to support it because (s)he would think (a) that his/her contribution would make little difference to the ultimate outcome and, (b) that his/her evasion would go undetected. As is obvious, in a small group, the above arguments lose their force. In the case of a fence between two neighbors, one's non-contribution would be glaringly apparent and would leave the fence at best only half constructed (assuming each neighbor was supposed to construct half the fence). Let us now look at Olson's model formally where i is a typical individual in the (potential) group: T : the level of public good C(T) : the cost of public good with C(0)>0; C'(T)>0; C''(T)>0. C(0)>0 implies that some level of organizational costs are incurred before any public good is attained. The assumptions on the derivatives give the standard U-shaped average cost curve. Sg : the value to the entire group of a unit of T. 33 Fi : the fraction of individual i1s valuation of T in Sg. Vg = SgT : total benefit to the group of T units of the public good. Vi = FiSgT : the benefit to i of T units of the public good. Ai = Vi(T) - C(T) : the advantage that i would obtain providing T units by himself/herself. The first-order condition (the second-order conditions are satisfied by the convexity of the cost function) for an individual optimum is: SA/ST = 0 which implies (1) FiSg = SC/ST marginal marginal cost benefit to i to group Let T* be the level of public good that satisfies (1). There is a second condition that must hold: (2) A, > 0 or FjSgT* > C(T*) 34 i.e., the net benefit to the individual must be positive. If the value of T* from (1) does not ensure satisfaction of (2) , the individual will not cooperate in the provision of the public good. One ignores the positive externality one's effort has on others. In evaluating one's course of action, one considers V., that is, the benefit to i alone and not Vg which is the total benefit to the group. If the public good in question has a positive externality, it is possible that = Vf (T) - C(T) could be negative but that Vg - C(T) may be positive. The first-order condition for a social optimum is: (3) Sg = SC/ST Let the level of public good that satisfies (3) be T**. Thus, group benefits are maximized at T** while individual benefits are maximized at T*. These results lead to two corollaries: [I] The small exploit the big. "The big" means Fi is large and thus F{ is more likely to be positive. "The small"— with low Fj— knows that the big will provide some level of T (though not necessarily T**) to which 35 (s)he will also have access. It is, therefore, rational for this this person to free ride. [II] The larger the number of identical people, the farther away the actual (realized) value of T will be from T**. Hence, rational individuals are unlikely to provide any, and at best, sub-optimal levels of, a public good in the absence of coercion or selective incentives.4 2.2.2 HARDIN Hardin refines Olson's analysis and adds to the latter's contribution. If A. > 0 for some i, the group is privileged and, if Af < 0 for all i, the group is latent. Group success is likely in the first case and unlikely in the second. In Olsonian terms, absolute latency, k, is the minimum for which there exists a T such that Ai > 0 Thus, k is the size of the smallest subgroup that benefits. 4Thus, for example, defense can be financed only by compulsory taxation. 36 Hardin makes clear the distinction between latency and size. A group can be either privileged or latent; with respect to size, it could be small, intermediate, or large. Olson appears to use the terms "large" and "latent" synonymously as he does "small" and "privileged". Hardin, on the other hand, argues that group failure is a function of latency, not size. Thus, a group is privileged if k = 1; it is latent if k > 1; it is intermediate if 1 < k < X, where X is "small" and will vary according to the context. When the public good is national defense, X is likely to be greater than when it is a community park. The essential difference between Olson and Hardin hinges on crowding. Olson, in effect, assumes that the public good suffers from crowding. Sg, the value to the entire group of a unit of the public good, is fixed. This means that, if any new member enters the group, the old members' Fi's must fall and thus group success becomes less likely. 37 When the members of a group have asymmetric interests— that is, when the value placed on a given unit of the good and/or its cost are not the same— prospects for collective action will, ceteris paribus, be enhanced. Hardin identifies four types of asymmetries: (1) nonfungibiltv of goods, or asymmetry between costs and benefits, that is, these are not monetary or readily exchangeable for money. (2) inequality of benefits, or asymmetry of demand: prospects for collective action may be enhanced or diminished. The two polar cases are analogous to Hirschman's [197 0] exit and voice responses. If there exist very intense demander(s), k will tend to be small (or equal to 1) and group success is likely (or certain). If, on the other hand, there is a private good substitute, the intense demanders may well opt for it. (3) heterogeneity. or asymmetry in the content of goods: members may desire the good for its various attributes and this enhances the prospects for cooperation. 38 (4) hysteresis. or asymmetry between gains and losses: the utility gained in obtaining a new good is less than that lost in giving it up. This may enhance group cooperation; an example would be the activity of environmental groups to obtain pollution control. Hardin also refers to contractarian provisions wherein success depends on both the existence of a clear consensus of what is fair and on extrarational motivations. In conclusion, it is important to note that the individual in Olson and Hardin is very similar. (S)he is an egoistic, utility-maximizing person devoid of social content. Not surprisingly — except for minor differences— their results are similar. 2.2.3 AXELROD Axelrod [1980a, 1980b, 1981, 1984] may be viewed as a unified body of work. In fact, Axelrod [1984] provides a consolidation and extension of his earlier works. Axelrod restates the free-rider problem in game-theoretic terms, similar to Figure 1.1. He 39 agrees with the standard result that, if two egoists played the game once, the dominant choice would be defection. The same holds true if the game is played a known, finite number of times. There is no incentive to cooperate on the last move, thus none on the next-to-last move,... and the entire process unravels to defection on the very first move. Note that this is true if, and only if, each player i is sure that the other j will defect. The above result is, however, not true when there is an unknown, finite number of iterations.5 Thus, 11. . . cooperation [emerges] ...[because] the players might meet again." Today's choices will affect later ones. The future "casts a shadow" on the present, that is, it influences the current strategic situation. 5But, as Kreps and Wilson [1982] have shown, a suspicion (hope?) by one player i that the other player j may adopt a TIT FOR TAT (TFT) strategy may induce i to also play TFT until (s)he encounters defection on the part of j. The uncertainty inherent in a situation of imperfect information may lead to cooperation and not automatic defection as the others claim. 40 If one cares sufficiently strongly about the future— that is, if the discount parameter w, 0 < w < 1, is "sufficiently" large— there is no best strategy independent of what the other player is doing. If one player always defects, so should the other. Under all other conditions, a TIT FOR TAT (TFT) strategy is shown to be superior to all other strategies. A TFT strategy is defined as cooperation on the first round and then mimicking the other player's action on the previous round in subsequent rounds. The results of two tournaments where contestants were asked to submit computer programs for the games are reported. TFT won both tournaments, even when before the second one all the programs entered in the first tournament (including the winning TFT program) were made known to all contestants. Furthermore, TFT was found to be robust, even in a simulated sequence of hypothetical future rounds of the tournament. Axelrod gives four reasons for TFT1s success: (a) niceness: the player always cooperates on the first round; (b) retaliation: defection by another 41 player is punished by defection; (c) forgiveness: if a player defects on one round but then cooperates, (s)he is forgiven, that is, the other player now cooperates with her/him; and (d) clarity: the above rules are simple and thus known to the players. Axelrod's major contribution lies in the fact that he considers strategic interaction situations of time horizons longer than that of the one-shot PD prototype but shorter than infinitely long. His assumption about an unknown, finite time horizon appears unobjectionable and his results are intuitive. The main problem is that he provides no reason why a player should cooperate on the first round. It is only sensible to do so if ones believes (knows?) that the other player is also likely to cooperate at least on subsequent rounds, if not on the first round. The success of TFT on purely mathematical grounds is clear but the explanation for its adoption by a rational, egoistic person is unclear. It appears imperative 42 that we examine the social context to help explain this paradox. 2.2.4 COLEMAN Coleman [1986a] addresses the question as to why there is no war of all against all a la Hobbes. He criticizes the sociological viewpoint that takes as its starting point a social system in which norms exist and people just observe these norms. Such a theory— though it sidetracks the Hobbesian war— is unable to handle situations where no norms exist. It is important to show not only how a collective decision is made but also how a situation of conflict ensues. Coleman [1986a] assumes an individual who is the polar opposite of the socialized individual: "unsocialized, entirely self-interested, not constrained by norms of a system, but only rationally calculating to further his own interests.1 1 He analyzes a legislative game where utility-maximizing actors exchange votes, that is, they give up votes on issues of little interest to them for others' votes on issues of greater 43 interest to them. It is necessary that there must be more than one issue to be decided upon; otherwise, as is obvious, no exchange is possible and the Hobbesian war will ensue. He finds that players neither quit, nor use external power, that is, power lying outside the interdependent actions of the given set of actors. They actively exchange power, seeking what is useful for what is not. Furthermore, as the number of sessions continued, unreliable actors endeavored to restore others' confidence in them. Long-term rationality required developing trustworthiness even at the expense of immediate gain. Coleman concludes that there is only one relevant sense in which the actors are socialized: they are socialized as rational actors, able to estimate the immediate and future consequences of their actions. They may be socialized in other ways but such considerations are not relevant to the game. A major aspect of socialization then is not the internalization of norms but the ability to foresee the long-term consequences of different 44 strategies. The result is not a passive, "socialized" individual but an even more calculating, rational one. Coleman does not deny the existence of norms. But he correctly points out that, to the extent that actors can withdraw, acquiescence to norms will only be given in terms of their own interests. Thus, a social theory should start not with norms and socialized individuals but with "collective actions and rational actors, each with interests and power relative to these actions."6 Coleman [1986b] appeals to sociologists to adopt methodological individualism. The behavior of the system is a consequence of the actions of its interdependent members. Thus, the purpose, goal, and homostatis of the actors should be studied and these should not be posited for the system. Purpose and goal are useful in theory construction only if they characterize elements of the system. What needs to be studied is how 6Coleman [1986a] 45 purposive, individual action leads to aggregate behavior and how the individual actions are in turn constrained by the behavior of the system. Economists refer to this micro-to-macro problem as the problem of aggregation. Methodological holism can be depicted in causal diagram form as: This is the Weberian argument that one macro variable (religion) affects another (economic system). Methodological individualism is depicted as : ( Protestant ) religious doctrine ( Capitalist ) economic system Figure 2.1 : Macro-level relation 46 Macro Level (Protestant) (Capitalist) religious doctrine economic systems i i i i I 2 x I3 . . ! 7 ! individual values orientations to economic behavior Figure 2.2: Macro-micro-macro relations It is the type-3 relation that provides the most challenge for both theoretical and empirical research. Weber7 showed the effect of Protestantism on individual values (type-2) and their effect on orientation to economic behavior (type-1). But, if the latter necessarily led to capitalism, Weber failed to show this. Coleman thus argues for concerted efforts to study these type-3 relations, while recognizing that these are the most difficult to analyze. Coleman [1987] examines the role of norms in a rational choice theory. Norms constitute constraints similar to any other type of 7Weber, Max. 1958. The Protestant Ethic and the Spirit of Capitalism. Scribner's, New York. 47 constraints. It may be useful to distinguish between the effects of two types of norms— something Coleman does not explicitly do. Internalized norms affect the utility functions of actors while norms requiring external sanctions could affect either the (expected) utility function or the constraint or both. Such factors as the utility differential attributable to observing the norm, the probability of being detected if one violates the norm, and the likelihood of sanctions being imposed if detected, all come into play. Coleman's central premise is that norms arise when there are externalities, either positive or negative, and when markets do not exist or transaction costs are too high. Thus, high school students give deference and prestige to winning athletes, and resent scholars, especially if the teacher grades on a curve. A winning athlete is a positive externality because (s)he brings pride and honor to the school. A scholar, on the other hand, is a negative externality to almost everyone because, if the teacher grades on a curve, it will 48 almost certainly mean lower grades for a typical student than if the scholar were not in the class. The acceptance of norms cannot, in all cases, be accounted for by rational choice theory. But, once the time horizon becomes expanded, the situation is different. Obedience to a norm may cause immediate disadvantage but may be beneficial from a long-term perspective. Moreover, norm observance leads a person to impose it on others, that is, the externalities of others' actions depend on one's own actions. Among a group of men where the norm is to wear a tie, one's reaction to a newcomer wearing a tie will depend on whether he himself is. In situations where those whose actions being sanctioned are themselves not sanctioners (children, for example), the acceptance of legitimacy is not explicable directly by rational choice. Coleman [1987] writes that, in such cases, norm acceptance is affected by "... the number of other persons holding the norm, and the numbers who do not accept it." It appears that 49 theories of critical mass would be most useful in these cases— something Coleman does not discuss. Finally, in the case of internalized norms, the claim of others' control is thought to be legitimate because the agent identifies it with his own interests. Social interactions can be viewed as exchanges between the parties; the interaction continues if the exchange is beneficial. Many such exchanges cause externalities, that is, they are public goods or bads. Social norms are expressions of the legitimate right to exercise control over these actions. They can lead to a socially efficient outcome because externalities are taken into account. Thus, social norms constitute social capital: they result in higher levels of satisfaction though perhaps at the cost of those whose actions are constrained. Coleman [1988] provides a logical continuation of Coleman [1987]. He assumes rationality but rejects the extreme individualism that is usually associated with it. Thus, he 50 argues, "persons’ actions are shaped, redirected, constrained by the social context; norms, interpersonal trust, social networks, and social organization are important in the functioning not only of the society but also of the economy." Granovetter's [1985] concept of embeddedness is viewed as an attempt to give a structure to social organizations and relations with history and continuity that cause it to have an independent effect on the economic system. The conception of rational action is superimposed on by social and institutional organization, either endogenously or exogenously. Coleman [1988] defines social capital thus: "they all consist of some aspect of social structures, and they facilitate certain actions of actors— whether persons or corporate actors— within the structure." It inheres in relations between actors; examples would be the trust among wholesale diamond traders in New York City and the cooperation observable among shopkeepers in a Cairo market. 51 Social capital is less tangible than other forms of capital because it exists in the relations among people. A group within which there is more trust and trustworthiness can achieve more than another where these attributes are less in evidence. Thus, the value of this concept of social capital is that it identifies certain aspects of the social structure by their functions. What can constitute capital resources for individuals? Coleman identifies three classes: [A] Obligations, Expectations, and Trustworthiness of Structures: Suppose A does something for B and trusts B to reciprocate in the future. There is an expectation in A and an obligation for B. These obligations can be thought of as credit slips and the analogy to financial capital is obvious. [B] Information Channels: Information can be obtained by using social relations that are maintained for other purposes. 52 [C] Norms and Effective Sanctions: An effective norm is a powerful, though sometimes fragile, form of social capital. Coleman next tries to identify certain social structures that facilitate some forms of social capital: [a] Closure of Social Networks: As noted above, externality is a necessary, but not sufficient, condition for effective norm formation. B B \ / \ / \ / \ / \ / \ / \ / \ / A I ! \ / \ / \ / \ / \ / \ / V A Figure 2.3 Figure 2.4 In an open structure as in Figure 2.3, A can continue to impose negative externalities on B or C or both because the latter have no relations with each other and thus cannot effectively 53 combine forces against A. In a system with closure, as in Figure 2.4, however, B and C can impose a collective sanction against A. Intergenerational closure may be depicted in the following diagrams where A and D denote parents and B and C denote children. E A D F A D _________ l _________ l__________ l_____________________I II I I II I I I I I I II I I I I I I II I l B C B C Figure 2.5 Figure 2.6 In the community depicted in Figure 2.6 there is intergenerational closure, that is, the parents' friends are the parents of their children's friends; in Figure 2.5, they are not. Thus, in the first community (that is, the one depicted in Figure 2.6), a school-age child loitering in the streets during school hours would be sanctioned by his friend's parent— a form of social capital not available to the second community. 54 Closure also increases trustworthiness because defection imposes negative externalities on others and, in a structure with closure, sanctions are more likely to be effective. [b] Appropriable Social Organization: An organization, once created for some purpose, can be used for others and is thus social capital available for use. Closure and appropriable social organizations provide social capital when there are multiplex relations, that is, persons are linked in more than one context (neighbor, fellow worker, fellow parent, coreligionist) as opposed to simplex relations where persons are linked through only one relation. Multiplex relations allow the resources of one relationship to be used in others. Coleman [1990] is an important and monumental work (it is 993 pages long) which consolidates and expands on his earlier works.8 He remains the ardent rational choice theorist and his concept of 8For an easily accessible summary and review, see Frank [1992]. 55 rationality is similar to that of economists. He differs, however, from neoclassical economists by emphasizing the social context in all behavior by self-interested individuals. Just as the need for double coincidence of wants in a barter exchange can be removed by the introduction of money, so the need for double coincidence of wants in social exchange can be eliminated by the use of social capital. A does a favor to B, who reciprocates by inducing C (by using his/her, that is, B's social capital) to do a favor to C. In the absence of such networks, social exchange tends to be inefficient in the same way that the lack of money would render many mutually beneficial exchanges impossible. He also criticizes the neoclassical economists' depiction of independent individuals devoid of any interaction and feedback with other members of the society. Coleman emphasizes the importance of social norms but— not surprisingly, given his strong proclivity toward methodological individualism— insists that such norms must be 56 understood as the result of the purposive actions of rational individuals. The demand for norms arises from the presence of externalities. The fact that a poorly maintained lawn may impose external costs to other homeowners may help explain the norm that encourages everyone to mow their lawns regularly. Coleman's important contribution is the recognition that the presence of externality is not sufficient for norm formation. The critical problem on the supply side is the free rider problem associated with the imposition of sanctions. In this regard, he emphasizes the importance of networks of personal relationships and social capital. Thus, people living in small communities, who interact regularly with each other, are better able to sanction norm violators than are city dwellers. 2.2.5 GRANOVETTER Granovetter [1985] criticizes both the under and over-socialized views of man. The former views the individual as atomized and devoid of any social context. The latter considers him/her to 57 be completely constrained by social relations to the point where (s)he is completely dependent on these relations for all his/her actions. Interestingly, Granovetter points out that both views "have in common a conception of action and decision carried out by atomized actors." In the under-socialized view, atomization results from narrow egoistic rationality while in the over-socialized view it results from the internalization of norms, conventions, and beliefs. Thus, the existing social context is actually unimportant in both scenarios. This irony is apparent in Hobbes's Leviathan where egoistic utility-maximizers ("under-socialized") meekly surrender their rights to the Leviathan ("over-socialized"). If social influence is considered an external force, ongoing social relations and structures are irrelevant. Once it is known how the individual is affected, his behavior is automatic. A more sophisticated approach is needed where the social context "not only shapes its members but also is 58 shaped by them, in part for their own strategic reasons."9 Voluntary cooperation is observed more often than is predicted by models of egoistic behavior and much more often than predicted by models of opportunism. Fraud and force may be averted due to clever institutional arrangements that make such behavior too costly. This is the under-socialized view where these arrangements are substitutes for trust, and since rational individuals would still endeavor to evade these arrangements, deceit and fraud are still likely. The other view— over-socilaized— recognizes a generalized morality, that is, some degree of trust is assumed to exist because institutional arrangements alone cannot stop fraud. In the absence of such morality, one would be wary of handing over a 2 0-dollar bill for gas worth five dollars, or no one would leave a tip in a roadside restaurant far from home. Granovetter argues that instances of behavior explainable by morality are ’emphasis added 59 rare, especially if large costs are involved. It is not likely that one would leave a very large tip at a restaurant to which one is unlikely to return even if the service was excellent. He stresses the embeddedness argument: the roles of concrete personal relations and of the structure of such relations in generating trust. One prefers to deal with individuals of known reputation; ideally, one wants to transact with someone with whom one has had a fruitful exchange in the past. The result of strategic behavior can be dramatically different when others' behavior can be counted on. There are thus more stampedes in a burning theatre than in a burning house, even taking into account the density of people. PD situations are overcome by the personal relations people sharing a house typically have. It is important to realize that rational individuals are less interested in another person's general reputation than in whether that person will cooperate with them, and this depends on any past dealings they or their friends/acquaintances have had with him/her. 60 Granovetter is careful that his theory is not reduced to functionalism. He argues, first, that embeddedness is less sweeping than either morality or institutions. Social relations are present in differing degrees in different spheres of economic life and thus fraud does occur. Secondly, social relations may, in many instances, be a necessary condition for trust but they are not sufficient; indeed, these relations can result in greater fraud and conflict than is possible in their absence. 2.2.6 HECHTER The first two chapters of Hechter [1983] propose a theory of group solidarity. He assumes methodological individualism to demonstrate that both normative and structural explanations for group behavior are unsatisfactory. He agrees that normative constraints exist but questions the causal status of these constraints. Moreover, the existence of a norm is not sufficient for compliance. It is not possible to explain the difference in "deviant" behavior across social categories. In structural explanations, social 61 relations narrow the individual's choice set. Individuals form groups on the basis of shared common interests. But, as Olson [1971] and Hardin [1982] have shown, rational individuals will free-ride in such situations. Hechter starts by defining a group as a set of individuals who interact with each other and can thus influence one another's actions.10 The stronger the relations that bind individuals in a group, the greater will be the group's influence on its members. Because there can be differences in the strength of social ties, solidarity must be thought of as a relationship of variable strength between actors that is analytically independent of their social roles. The existence of a moral standard that members act according to some collective standards of conduct or norms is a distinguishing feature of a solidary group. Individuals who merely have attributes in common do not have this obligation to act in a prescribed 10It is the interaction which provides members the means to influence each other's actions. 62 manner. Groups thus differ both with respect to the obligations imposed on members and the latters' compliance. Thus, "[t]he more compliance a group demands of its members, and can obtain, the higher its solidarity." Compliance to group norms is sought even when it is against an individual's self-interest. A rational actor will choose the action that furthers his/her own goals. Thus, when there is a conflict of interest between group and individual goals, group obligation is best viewed as a tax. The greater one's dependence on the group, the more likely one is to comply with its norms. Furthermore, compliance can be ensured only when there is proper monitoring. Hechter thus identifies two separate determinants of variations in the individual's compliance with group norms: (1) differences in his/her dependence on the group and (2) differences in the group's monitoring capacity. The first depends on the intensity of the individual's preference for the good(s) provided by the group and, on the supply side, on the 63 availability of substitutes. The second depends on the group's metering and sanctionong activities. Hechter [1984] is, as the title suggests,1 1 a detailed consideration of monitoring, the costs of which are often underestimated. Any monitoring system has metering, sanctioning, and allocation capabilities. Metering capacity depends on information about individual compliance with group obligations; sanctioning capacity on resources available for sanctions; and allocative capacity on the group's ability to distribute these sanctions to reward/punish compliance or non-comp1iance. He concludes that there is no unique solution to the problem of minimizing monitoring costs. There are costs involved in each of the above three functions and these costs constitute the 11The article is entitled "When Actors Comply: Monitoring Costs and the Production of Social Order". The complete citation will be found in the bibliography at the end of this chapter. 64 major constraints to the adoption of any monitoring system. Given that monitoring is costly, how can compliance be produced and maintained in a group of egoistic individuals? Four possible solutions are mentioned: (a) to pass the monitoring costs on to consumers or clients? (b) to adopt institutional arrangements such that members have incentives to meter each other; (c) to apply public sanctions; and (d) to threaten expulsion or ostracism. Depending on the situation, one or more of the above may be effective. None, however, does away with monitoring. Thus, compliance is unlikely in large and/or dispersed groups. It appears, however, that compliance is more prevalent than rational choice theory would predict. Arrow [1970] concludes that employees follow orders and citizens obey the law to a much greater extent than can be explained by control mechanisms alone. If sufficient compliance exists in the absence of monitoring— altruism, 65 internalized norms, for example— research on determinants of preference formation is crucial. Hechter [1987] is an extension of Hechter [1983]. A group's solidarity depends on the extensiveness of its obligation (a) and the degree of member compliance (b): Solidarity = f(ab) The greater the average proportion of each member's private resources that is devoted for collective purposes, the greater is the solidarity of the group. In addition to the normative and structuralist perspectives described above, Hechter also considers the functionalist perspective and concludes that none of these can explain group solidarity. First, they are incapable of explaining differences in the extensiveness of group obligations and, secondly, none explains when members will comply with these obligations. Rational choice theory would predict that the more public a collective good is, the less likely it is to be produced at optimal levels, if it is produced at all. Olson's solutions included coercion and selective incentives, social or material. These solutions lead to additional questions. One relates to the problem of determining just how the state attains (and retains) its coercive ability in rational choice theory. The second is that selective incentives, although endogenous, are likely to work only in small groups. Material selective incentives are more fungible and less likely to depend directly on group size. The selective incentive argument raises even more fundamental questions at both the theoretical and empirical levels. If selective incentives are required to produce some collective good, then they (incentives) are merely another type of (potentially) costly public good whose production is not certain. In particular, following Olson, unless the incentive is of greater value than the individual's share of the cost of the collective good, it is irrational to contribute. The crucial question then is: "[u]nder what conditions will sufficiently valuable selective incentives be produced?" The 67 empirical difficulty with this argument is that, if incentives are necessary and sufficient for public good provision, group survival should be independent of the good desired. The longest-lived groups should be the ones that provide the most incentives, and the determinants of collective good production should reflect the market in incentives. But, in this case, it is hard to explain how disadvantaged groups (early labor unions, for example) could mobilize against more powerful antagonists (the employers). Rational choice theory, on the other hand, provides both a reasonable explanation for an individual's desire to form and participate in groups and testable hypotheses about their size. Thus, Buchanan's [1965] theory of clubs would imply that group size will be a function of both crowding and cost-sharing arrangements. The theory of group solidarity presented here is basically identical to that in Hechter [1983]. Several additional factors are considered as determinants of dependence: (i) transfer costs or entry and exit costs and (ii) the strength of 68 personal ties. Constitutional and legal arrangements are crucial in generating dependence. Whenever rights are delegated to a group rather than to individuals, dependence is increased: examples include the caste system in India and blacks in South Africa. Dependence may be eroded by the welfare state since the individual now enjoys benefits that were previously supplied by churches, trade unions and others. 2.2.7 HECKATHORN Cooperation is neither rare nor fragile even when it is not anticipated. Axelrod [1984] showed that informal truces during World War I arrived at by ordinary soldiers can be viewed as cooperation in an iterated PD situation. In such regimes of decentralized control, the robustness of cooperation generally involves systems of reciprocity wherein one's cooperation is conditional on that of others'. Heckathorn [1988] analyzes systems of collective sanctions (CS) where an individual's violation of, or compliance with, a rule or norm leads to punishment or reward by an external agent of all members of the group. Instances of CS include the burning of entire villages for a sniper's actions by the Pakistani army in Bangladesh, tribal/kinship responsibility for acts of its members, and the blowing up of homes of suspected Palestanian activists by the Israeli army. Social scientists have viewed norm emergence as depending on: (i) inclinations: actor's own preferences; (ii) regulatory interests: actors' preferences about others' behavior; and (iii) enforcement resources: measures for enforcing norms. This paper by Heckathorn focuses on CS norms whose regulatory interests derive from collective sanctions. When collective sanctions are imposed, each player has three choices: (i) cooperate; (ii) defect; or (iii) rebel, that is, refuse to play. It is found that a variety of responses may be rational when a group is subjected to collective sanctions. Exogenous compliance norms may be created as was arguably the case of Norway under Nazi occupation. CS systems may lead to 70 passivity (Russian occupation of Czechoslovakia, 1968-89) or to revolt (Bangladesh in 1971). The effect of control depends sensitively on both the attributes of the controlling agent (monitoring capabilities, strength of sanctions, and vulnerability to revolt) and those of the group (group size, degree of intragroup control, proportion of potential violators, and cost of normative control and revolt). Heckathorn [1989] analyzes the second-order free-rider problem which arises because the sanctioning system itself is a public good. The first-order free-rider problem is participating in the production of the public good. Actors face four choices: (1) full cooperation (first- and second-order cooperation); (2) private cooperation (first-order cooperation and second-order defection); (3) hypocritical cooperation (first-order defection and second-order cooperation); and (4) full defection (free riding in both instances). He argues that second-order cooperation is more robust than first-order cooperation, that is, 71 it flourishes in conditions where the latter fails.12 The analogy to Matza and Sykes' [1961] "public morality" and "subterranean values" and to Kuran's [1987] "private preferences" and "public preferences" is obvious.13 When cooperation is less than perfect, it tends to be of the hypocritical type rather than the private type. An ethic of hard work, absolute fidelity, and scrupulous honesty may be espoused to subordinates but, in the company of friends, such principles may be derided as excessively moralistic. The recent behavior of TV evangelists also fits this model. Thus, hypocritical behavior— though often maligned— can, under some circumstances, help create what Axelrod [1986] calls "metanorms" which are norms mandating participation in norm enforcement. 12Ullmann-Margalit [1977] demonstrated numerous examples of PD norms in the many societies she studied. Moreover, sanctioning systems are prevalent even when their effectiveness appears fairly limited. 13The "dual value system", in Kuran and Matza and Sykes, implies in a PD (or PD-type) situation that second-order cooperation is more robust than first-order cooperation. 72 2.2.8 MAYNARD SMITH Evolutionary models, which are pervasive in the natural sciences, are conspicuous by their absence in the social sciences. Darwinian theory has been almost exclusively been the purview of the natural sciences even though, as Basu [1991] points out, Darwin had been influenced by Malthus. The pioneering work of Maynard Smith and others represents a large step forward in rectifying this void. Their work combines evolution with game theory and allows for strategic interaction between players. The seminal work is that of Maynard Smith and Price [1973]. The question they seek to address is why conflicts between animals of the same species do not usually lead to "total war" causing serious injury but are of "limited war" variety. In many snake species the males wrestle each other without using their fangs. If one recognizes the fact that the winner in such contests gains mates, dominance rights and territorial choice which will tend to transmit its genes to future generations at higher frequencies than those of the loser's, 73 one may logically expect natural selection to lead to "war until death". But what is prevalent in the animal world is "limited war" using inefficient weapons or ritualized tactics unlikely to cause serious injury to either contestant. The prevalent explanation for the observed "limited war" was the operation of "group selection"; it was felt that otherwise many individuals would be injured and the survival of the species would be in jeopardy. Maynard Smith and Price [1973] reject the concept of group selection in most cases and state that there must be "individual selection" for limited warfare in so many species. This means that limited warfare "must be differentially advantageous for individuals." The group selection theory would appear to be a functionalist explanation and furthermore it assumes that actions desirable for the group as a whole would be undertaken. This would be in contradiction to Olson's observation that group satisfaction is not necessarily maximized by individual behavior. 74 The concept of an "evolutionarily stable strategy" (ESS) is fundamental to the formal models of conflict resolution where strategies are chosen under individual selection. An ESS "is a strategy such that, if most of the members of a population adopt it, there is no "mutant" strategy that would give higher reproductive fitness." A computer simulation is attempted where there are two types of conflict tactics: (i) conventional (C) which are unlikely to cause serious injury and (ii) dangerous (D) which are likely to cause serious injury if employed too long. In addition, a contestant may retreat (R). A "strategy" is a set of rules that gives probabilities to C, D, and R as functions of what has previously occured in the current contest. Five possible strategies were programmed: (1) Mouse (M): Never plays D. Plays R against opponent's D. Otherwise always plays C until the contest has lasted a preassigned number of moves. (2) Hawk (H): Always plays D. Continues the contest until seriously injured or opponent retreats. 75 (3) Bullv (B): Plays D on the first move. Plays D against C and C against D. Plays R if opponent plays D a second time. (4) Retaliator (RT); Plays C on the first move. If opponent plays c, plays C but plays R if contest has lasted a preassigned number of moves. If opponent plays D, retaliates with a high probability by playing D. (5) Prober-Retaliator (PR); If making the first move, or after opponent has played C, with high probability plays C and with low probability plays D but plays R if contest has lasted a preassigned number of moves. After giving a probe, reverts to C if opponent retaliates but takes advantage by continuing to play D if opponent plays C. After receiving a probe, plays D with high probability. The H strategy is "total war" while the others are variations of "limited war" and computer simulation was employed to see which type of war strategy individual selection favored. A set of assumptions was made regarding the probability: (i) of serious injury from a single D play (.1), (ii) that PR will probe on the opening 76 move or after opponent plays C (.05) and (iii) that RT or RT will retaliate against a probe (1). Assumptions were also made regarding payoffs for winning (+60), suffering serious injury (-100) and so on. The obtained: following average payoffs were Table 1: Average Payoffs for Strategies Five Different OPPONENT M H B RT RP CONTESTANT M 29.0 19. 5 19. 5 29.0 17.2 H 80.0 -19. 5 74.6 -18. 1 -18.9 RECEIVING B 80. 0 4.9 41.5 11.9 11.2 RT 29. 0 -22. 3 57.1 29.0 23.1 PAYOFF RP 56.7 -20.1 59.4 26.9 21.9 To examine whether a strategy is ESS, it is necessary that it be the most profitable strategy in a population almost entirely of the same species, that is, a contestant will almost always 77 have an opponent of the given species. Examining the column H in Table 1, M and B are more successful than H. Natural selection will thus cause M and B behavior to increase and H to decrease; thus H is not ESS. RT is an ESS since no other strategy does better though M does equally well. RP is almost an ESS since only RT does better. The most important result is that "limited war" is shown to be superior under individual selection to "total war" (H). The analogy to economic and social behavior is the lack (or limited amount of) conflict that we observe and that a Hobbesian sovereign is not required to help explain this. A further point to note is the basic similarity between RT (or RP) and Axelrod's TIT FOR TAT (TFT) strategy. Thus, a stable strategy requires that "contestants should respond to an "escalated" attack by escalating in return." The population would come to consist mainly of RT or RP with the other strategies occurring with low frequencies. The balance between RT and RP will 78 depend on the frequency of M because probing is an advantage only against M. Maynard Smith [1974] is an extension of Maynard Smith and Price [1973]. The latter work, described above, proposed a model of conflict resolution at the individual level where the success of any strategy depends on the strategies undertaken by others. The former work considers two important issues: (i) the assignment of utilities and (ii) the nature of the ESS. The application of game theory to human conflicts requires that numerical values (or utilities) be assigned to the possible outcomes. Is it justifiable to put utilities from financial gain or loss on the same numerical scale as those from injury or death? In human conflicts, strategies are chosen to maximize the satisfaction of wants while in animal contests they are selected to maximize the fitness of the contestants. Thus, "[t]his equivalence between utility and contribution to fitness is the main justification for applying game theory to animal contests." 79 A formal definition of ESS is provided: a strategy I is an ESS if the expected utility of I played against itself is greater than the utility of any other strategy J played against I, that is, (1) E, (I) > E, (J) where E is the expected utility of the strategy in parenthesis played against that indicated by the subscript. In a population consisting mostly of individuals adopting I, rare mutants which adopted J would not increase in frequency. The population would be stable under mutation and selection. What happens if Ej(I) = Ej(J)? I is still an equilibrium strategy but it need not be stable. In a population of which a fraction p adopt I and (1-p) adopt J, the expected "fitnesses" are: (2) E(I) = pEj (I) + (l-p)Ej(I) (3) E(J) = pEj (J) + (l-pJE^J) For all p < 1, E (I) > E (J) or I to be evolutionarily stable requires: (4) Ej (I) > Ej (J) Thus, I is an ESS if, for all alternative strategies J, either (5) Ej (I) > Ej (J) 80 or, (6) Ej (I) = Ej (J) and (7) Ej(I) > E^J). There may be more than one ESS for a given game, that is, the stability need not be global. The population would evolve to a different ESS according to its initial conditions. As shown in Table 1 above, the balance between RT and RP would depend on the frequency of M. Maynard Smith [1982] introduces a method for modelling evolution but does not, in general, deal with any specific problem to which the method can be applied. For our purposes, the most interesting part of the book is Chapter 13 entitled "The Evolution of Cooperation". He states at the outset that cooperation among animals presents problems for Darwinists unless there are indirect benefits to the cooperators. If, on the other hand, some property of one species exists solely to ensure the survival of another, Darwinian theory would be disproved. In a Prisoner's Dilemma (PD), if each individual plays only once, the only ESS is to defect. As Axelrod [1981] has shown, cooperaton evolves if individuals play against the same 81 opponent. He found that TIT FOR TAT (TFT) was the winning strategy and was thus an ESS. The reasons TFT is ESS are: it is "nice" (it is never the first to defect); it is "provokable" (defects at once in response to defection; and it is "forgiving" (readily returns to cooperation if its opponent does so. Axelrod proves that TFT is stable against invasion by any other strategy provided the sequence of games against each opponent is long enough. But "always defect" (D) is also an ESS. We observe, however, cooperative social behavior in interactions among family, kin and tribe. Why is the other ESS (D) not commonly observed? Suppose individuals play mainly against neighbors and that there is some clustering of individuals with similar strategies. TFT, even when rare, can invade D because TFT does well against players using the same strategy and clustering does not enable D to invade TFT. Thus, family or kinship is needed for the initial spread of a trait (such as a norm or belief) which, once established, can be maintained by mutualistic effects alone. 82 If cooperative behavior is stable, successful individuals must pass their characteristics on to more descendants than those who are not. Maynard Smith provides three hereditary mechanisms: (1) Genetic: Differences between individuals adopting different strategies are in part genetic, that is, caused by differences between the fertilized eggs from which they developed. (2) Learning: If almost everyone were playing TFT, it would not pay anyone to adopt any other strategy. This may be learnt either in practice or in imagination. (3) Cultural Inheritance: If individuals acquire their behavior by learning or imitation and if they are more likely to copy successful mentors, the result would be an increase in behavior patterns like cooperation which are stable. Thus, norms which lead to increased utility are more likely to persist than those that do not. 2.2.9 BASU Basu [1991] seeks to examine the relation between social norms and evolutionary processes. For a social norm or civil institution to persist, 83 two conditions must hold: (i) it is individually rational to adhere to the norm once it exists and (ii) it is robust in terms of evolutionary processes. Basu proposes what he calls "minimal functionalism", that is, if a society has survived for a long period of time with certain norms, these norms cannot be too harmful for the people. Norms may render some strategies out of bounds, e.g., picking pockets may not be a reasonable career for some people, regardless of the monetary payoff.14 A civil norm is a subset T of all conceivable strategies S. The agent chooses the optimal strategy from within T and considers (S-T) outside his/her realm of choice. A civil norm is evolutionarily stable if it is immune to invasions of small numbers of agents with other civil norms. Basu's main contribution is that he points out that to ensure that a norm is uninvadable by all conceivable norms is to err on the side of 14Not breaking a queue, Basu points out, may be a norm in Finland. A mutant (migrant from another country) who does not share this norm could presumably exploit this situation. 84 caution. If a mutant in one society is a member of another, (s)he must have a norm which is itself stable. Thus, it needs to be studied whether the society's norm is uninvadable by other stable norms. Since this introduces recursion in the definition of a stable norm, Basu uses a set valued solution concept. He defines a conclave of evolutionarily stable strategies (CESS) which can incorporate non-Nash equilibria as opposed to ESS which must be a Nash equilibrium. He provides reasonable numerical examples of CESS which are non-Nash equilibria. Although the set-valued solution technique is novel, Basu appears to take too static a view of norms. He distinguishes— correctly, in my view— between "teachings" and "common teachings". The former is what we teach our children: for example, don't go to the balcony because the railings are broken. The latter is something we teach our children, teach them to teach their children, and so on. Thus, a common teaching is "social science's counterpart of the gene in biology." He also distinguishes between progressive and regressive civil norms which are, respectively, norms which aid the progress of its adherents and those that do not.15 Basu appeals to exogenous change to lead to the demise of existing norms and usher in new ones. He views this as an evolutionary process similar to that in England where, as industrialization increased, darker moths became the dominant variety because they blended more easily into the soot laden background and were thus better protected from their predators. Exogenous changes in society could thus change the civil norms and provide the basis for studying the dynamics of civil institutions. 15This means that norms could either expand or shrink the choice set in the future. It is incorrect to view the presence of norms as necessarily restricting the (future) choice set. A norm against stealing could facilitate growth while one against trading could hinder it. 86 2•3 Prisoners1 Dilemma or Assurance Game? The above citations ranged from the pessimistic views16 of Olson and Hardin on theoretical grounds to the more optimistic views of Coleman, Hechter and Heckathorn. The basic issue to be addressed is whether PD is a realistic depiction of the choices facing society. It is to be noted that since neither Olson nor Hardin consider norms and since their agents are selfish utility-maximizers, the PD is relevant to their scenarios and, not surprisingly, cooperation is not achieved except under special conditions such as the presence of coersion or by-products. It is doubtless true that, as noted above, that we observe many more instances of cooperation than would be predicted in a pure PD scenario. To resolve this paradox, we turn to the Assurance Game (AG), first set forth by Sen [1967].17 He assumes a community of N 16with respect to the likelihood of successful group formation 17Details of the difference in payoff matrices of PD and AG are provided in Chapter 3. individuals with two choices X and Y. The preference ordering satisfies two conditions: (1) given what others do, the individual prefers X and (2) given the choice between everyone doing X and everyone doing Y, (s)he prefers the latter.18 This implies the following results: (a) The outcome is Pareto-inferior, that is, the outcome is regarded as strictly worse than the alternative by everyone. (b) Irrespective of each person's expectations of the others' actions, each does X; this atomistic result is devoid of any interdependence. (c) The policy of everyone doing Y, even if agreed upon, could come about only through compulsory enforcement. Consider a somewhat different preference ordering where (2) remains the same but (1) is modified. Now, in the special case where everyone else does Y, the individual prefers to also do Y. 18It is clear that when N=2, these two conditions imply PD. See Figure 1.1 where it is better for A to defect regardless of what B does but given a choice between both A and B defecting and both cooperating, A prefers the latter. 88 Note that result (b) no longer holds; expectations about the others1 behavior must now be taken into account. If it is expected that others will do Y, so will the typical individual. Result (c) does not hold either because, in the case of complete assurance, it is not in the person's interest not to do Y. In a society consisting of people with history, culture an norms, it appears reasonable that expectations about each other's behavior play a role in influencing the typical individual's behavior. The AG (or PD-type) situation is more realistic on empirical grounds than the pure PD.19 As shown in the model in the next chapter, these expectations in interdependent decision making situations may form the key to the proper understanding of group success and failure. 19See Chapter 3 for more details. 89 BIBLIOGRAPHY Arrow, Kenneth. 1970. The Limits of Organization, Norton, New York. Axelrod, Robert. 1980a. "Effective Choice in the Prisoner's Dilemma." Journal of Conflict Resolution 24: 3-25. .----------# 1980b. "More Effective Choice in the Prisoner's Dilemma." Journal of Conflict Resolution 24: 379-403. --------------- . 1981. "The Emergence of Cooperation among Egotists", American Political Science Review 75: 306-318. --------------- . 1984. The Evolution of Cooperation, Basic Books Inc., New York. ------- --------. 1986. "An Evolutionary Approach to Norms." American Political Science Review 80: 1095-1111. Basu, Kaushik. 1991. "Civil Norms and Evolution." unpublished paper. Buchanan, James M. 1965. "An Economic Theory of Clubs." Economica, 32 (12 5), new series: 1- 14. Coleman, James S. 1986a. Individual Interests and Collective Action,. Cambridge University Press. ----------------- . 1986b. "Social Theory, Social Research, and a Theory of Action". American Journal of Sociology 91: 1309-1335. ----------------- . 1987. "Norms as Social Capital". in G. Radnitzky and P. Bernholz (eds.), Economic Imperialism: The Economic Approach Applied Outside the Field of Economics. Paragon Press, New York, pp 133-155. 90 ----------------- . 1988. "Social Capital in the Creation of Human Capital", American Journal of Sociology 94 Supplement: S95-12 0. ----------------- .1990. Foundations of Social Theory. Harvard University Press, Cambridge, Mass. and London. Easterlin, Richard A. 1973. "Does money buy happiness?". The Public Interest. Number 30, Winter 1973. -------------. 1978. "New Directions for the Economics of Fertility", in Milton J. Yinger and Stephen J. Cutler (eds.), Major Social Issues: A Multidisciplinary View. The Free Press, New York. ------------- and Eileen M. Crimmins. 1991. "Private Materialism, Personal Self- Fulfillment, Family Life, and Public Interest". Public Opinion Quarterly, Volume 55: 499-533. Frank, Robert H. 1992. "Melding Sociolgy and Economics: James Coleman's Foundations of Social Theory". Journal of Economic Literature, Vol. XXX, No. 1, pp. 147-170. Granovetter, Mark. 1985. "Economic Action and Social Structure", American Journal of Sociology 91: 481-510. Hardin, Russell. 1982. Collective Action, Johns Hopkins University Press, Baltimore. Hechter, Michael (ed.). 1983. The Microfoundations of Macrosociology, Temple University Press, Philadelphia. --------------. 1984. "When Actors Comply: Monitoring Costs and the Production of Social Order", Acta Sociologica 3: 161-83. 91 ------------------. 1987. Principles of Group Solidarity, University of California Press, Berkeley, Los Angeles and London. Heckathorn, Douglas D. 1988. "Collective Sanctions and the Creation of Prisoner's Dilemma Norms", American Journal of Sociology 94: 535-62. ----------------------- . 1989. "Collective Action and the Second Order Free Rider Problem", Rationality and Society, Vol. 1, No. 1, pp 78-100. Hirschman, Albert. 197 0. Exit, Voice and Loyalty. Harvard University Press, Cambridge, Mass. Kreps, D. M. and R. Wilson. 1982. "Reputation and Imperfect Information", Journal of Economic Theory, Vol. 27, No. 2, pp 245-252. Kuran, Timur. 1987. "Preference Falsification, Policy Continuity and Collective Conservatism", Economic Journal 97: 642-665. Matza, David and Gresham Sykes. 1961. "Juvenile Delinquency and Subterranean Values", American Sociological Review 26: 712-719. Maynard Smith, J. and G. R. Price. 1973. "The Logic of Animal Conflict", Nature, London, 246: 15-18. Maynard Smith, J. 1974. "The Theory of Games and the Evolution of Animal Conflicts", Journal of Theoretical Biology, 47: 209-221. --------------- . 1982. Evolution and the Theory of Games. Cambridge University Press, Cambridge. Olson, Mancur. 1965. The Logic of Collective Action. Harvard University Press, Cambridge, Mass. 92 Sen, Amartya K. 1967. "Isolation, Assurance and the Social Rate of Discount", Quarterly Journal of Economics, Vol. 81, No. 1, pp 113-124. CHAPTER 3 A FORMAL MODEL OF GROUP BEHAVIOR 3.1 Universal Defection? As examined in Chapter 1, a PD situation should lead to almost universal defection. But, as Hechathorn [1988] observes, "[i]n the real world, cooperation is neither rare nor fragile." This is in sharp contrast to PD situations where, even in the unlikely event that cooperation occured, it would be likely to be fragile. Rapoport [1983] showed that defection is highly likely in the absence of preplay communication. It should, however, be remembered that such communication may be a necessary condition for cooperation but it is not sufficient. Each player has to have a certain amount of trust in the other(s). The typical solution to a PD situation is thus recourse to an external, cencentralized authority such as Hobbes's sovereign. What needs explanation is why cooperation is frequently observed in PD or PD-type (in the sense described below) situations even in the absence of a centralized authority. As was mentioned in 94 Chapter 1, informal social groups in New York can transact millions of dollars in the wholesale diamond market without any formal record-keeping and members of rotating-credit associations in India continue to contribute even after they receive an early payout. Not just cooperation, but sustained cooperation, is observed under apparently impossible circumstances. Axelrod [1984] showed that, during World War I, informal truces were arrived at by ordinary soldiers, often to the dismay of the officers. A virtual "live and let live" policy ruled in the trenches. If one side did not shoot, neither did the other. In the event that one side broke this implicit agreement, the other side would retaliate with even greater force. 3.2 The Restaurant Re-visited In order to understand phenomena of this kind, it is necessary to realize that most social interactions are not one-shot PD games but have unknown, finite (but large) numbers of iterations. People cooperate because they might meet the same people again in the future. Most of us would leave a larger tip at a restaurant we expect to revisit than at one we do not, 95 ceteris paribus. This does not mean that there may not be someone who would leave a large tip at a restaurant where (s)he has enjoyed excellent service even though (s)he is unlikely to return there: such behavior is just less likely in most, perhaps all, societies. This forms the crux of the model being developed: while various instruments of control— internal and external— exist that influence human behavior, it is one's private interest that ultimately motivates action (or inaction). How then does one explain the large tip left at a restaurant one is unlikely to return to? Such behavior may be motivated by: (i) the person's subjective ideals of fairness or justice or (ii) prevailing norms or beliefs or (iii) both. Situation (i) may be thought of as internalized norm or belief while (ii) is an external one. In the present example, (ii) is unlikely to be very significant because monitoring is not easy under these circumstances and thus defection is relatively costless. One can, however, think of other examples where, even though the probability of detection is low, the costs of noncooperation are so high that it is not 96 worthwhile to defect because the expected costs are too high. The bottom line is this: each person tries to maximize his/her expected utility. If (s)he knows that (s)he will be eating regularly at a restaurant and that a large tip will ensure (or make more likely) good service, (s)he is more likely to leave a large tip. On the other hand, one would leave a large tip regardless of the frequency of one's visits and dependent only on the quality of service if, otherwise, one would suffer psychic distress. Both are essentially attempts at maximizing utility. The first affects the budget constraint while the second affects the utility function itself.1 The robustness of cooperation can be understood if one recognizes the existence of reciprocity in human behavior. In a PD-type situation of multiple ^s noted in Chapter 1, the emphasis of rationality, it seems to me, should be on self- interested individuals, not necessarily selfish ones. In standard microeconomic analysis, this poses no problems. If the price of bread goes down, the quantity demanded goes up for most people, ceteris paribus. The selfish may then eat the bread himself/herself while the altruist may distribute it to the homeless. 97 iterations, it is reasonable to assume that actors are acutely aware of each other's interdependence. In a one-shot PD-type game, such information and even a solemn agreement may not be sufficient to render defection unlikely. But in a multiple-shot game, cooperation is a priori more likely. The restaurant waiter who is habitually sloppy and surly— that is, a noncooperator— is less likely to be tipped even by a person who leaves a large tip because (s)he thinks it is the correct thing to do. The person whose tipping is a function of quality of service would, of course, not leave a large tip the very first time. As mentioned in Chapter 1, it is the first round interaction which is crucial and largely overlooked in the literature. To continue with our example: to whom is a restaurant waiter likely to render good service the first time around? At least two answers are possible. He may be influenced by the patron's clothes and thus the probability (or lack thereof) of a large tip. The patron may be a friend of the owner and, in this case, the primary motivation may not be a large tip but continued employment and/or promotion. In both cases, the waiter is maximizing his expected utility 98 with the time horizon being much longer in the second scenario. In addition, there may be waiters who render good service because they believe it is a job requirement or are altruistic and are eager to help hungry and weary travelers. Even in these cases they are maximizing their (expected) utilities. To sum up, cooperation is more likely whenever the short-run gains from defection are less than the long-run gains from cooperation. Thus, Hechathorn argues [1988], "[c]ooperation tends to be more stable in groups of moderate size and reasonable cohesion." The analogy would be a smal1-town restaurant frequented almost exclusively by the local townsfolk. In such a situation, the waiter and the patrons have excellent information about each other's likely behavior. 3.3 Prisoners' Dilemma versus Assurance Game The existence of these social controls— internalized norms or beliefs, external sanctions, laws, customs— imply that the pure PD situation as depicted in Figure 1.1 does not hold. The pure PD situation is described by: (i) T > R > P > S and 99 (ii) R > ( T + S )/2 where R is the reward for mutual cooperation, T is the defector’s (that is, traitor's) payoff, S is the sucker's payoff, and P is the punishment for mutual defection. The crucial implication of (ii) is that an equal probability of exploiting and getting exploited is not preferable to mutual cooperation. For purposes of clarity of exposition, Figure 1.1 is reproduced below: Person j Cooperate Defect R=3 S=—2 Cooperate R=3 T=5 T=5 P=-l Defect S=-2 P=-l Figure 3.1 The above payoff matrix depicts a situation devoid of any form of social control. In the real world, however, if people do not cooperate, they may be subject to external sanctions, e.g., social ostracism. They may alternatively suffer psychic distress due to 100 internalization of norms. In such a situation, best described by an assurance game (AG), the crucial difference is that the payoff obtained by free-riding is less than that obtained by cooperating, that is, R > T. There is then what Platteau [1992] calls a "predisposition toward cooperation in the sense that (actors) derive a positive utility from cooperation as such." It needs to be pointed out, however, that actors are not altruistic; they do not like to be exploited by free riders, that is, P > S. Compared to the PD situation, the AG (referred to earlier as PD- type) can be depicted as: Person j Cooperate Defect R=3 S=—2 Cooperate R=3 T=0 Person i T=0 P=—1.5 Defect S=—2 P=—1.5 Figure 3.2 101 The crucial difference between Figure 3.1 and Figure 3.2 is that in Figure 3.2 condition (i) no longer holds. We now have R > T > P > S. Why should R > T hold? It does not seem sufficient to attribute this to a "predisposition toward cooperation" as Platteau [1992] does but to endeavor to fathom why this is so. There are at least two possible reasons: (a) the defector (traitor) knows that there is a positive probability that (s)he may be caught and punished on subsequent iterations and (b) internalized norms may lead to feelings of guilt. Thus-, although being a traitor is better than being a sucker (0 > -2), it is not better than being mutual cooperators. We now have two Nash equilibria (C,C) and (D,D). (C,C) is the better but riskier (since 3 > 0 but -2 < -1.5 ). The Pareto efficient solution (C,C) will not obtain if the agents are sufficiently risk averse. Even if communication between players takes place, cooperation can be achieved if, and only if, each player is convinced that the other would also cooperate. Thus, the AG does not mean that cooperation is certain, merely that it is more likely than in a pure PD situation where it is impossible. 102 Given our assumption of utility-maximization, each player attempts to maximize his/her (expected) utility. Suppose i expects j to cooperate with probability q and to defect with probability (1-q). Then i's expected utilities from cooperation (C) and defection (D) are: E U(C) = 3 (q) + (-2) (1-q) = 5q - 2 E U (D) = 0 (q) + (-1.5) (1-q) = 1.5q - 1.5 Assuming risk-neutrality, it is worthwhile for i to cooperate iff: E U(C) > E U(D) that is, 5q - 2 > 1.5q - 1.5; that is, q > 1/7. Of course, symmetric results hold for j, if we assume that i and j expect each other to cooperate with the same probability q. As long as each person knows that the other(s) will cooperate with a probability of at least 1/7, the Pareto optimum outcome (C,C) will obtain. I shall henceforth refer to this probability as the probability threshold and to the payoff matrix 103 depicted in Figure 3.2 as an AG. Note that the probability threshold is inversely related to R2, that is, if the payoff from cooperating is high, the degree of trust required will be low, ceteris paribus. In the situation depicted in Figure 3.1 where there was no social control, i's expected utilities from cooperation (C) and defection (D) were: E U(C) = 3 (q) + (-2) (1-q) = 5q - 2 E U(D) = 5 (q) + (-1) (1-q) = 6q - 1 Again assuming risk-neutrality, it is worthwhile for i to cooperate iff: E U(C) > E U(D) that is, iff 5q — 2 > 6q — 1 For any q between 0 and 1, the above condition cannot hold and thus we would expect universal defection and this is exactly the same conclusion that traditional PD analyses arrived at. As noted above, the differences between the payoffs in Figure 3.1 and those in Figure 3.2 involved T (the traitor's payoff) and P (the punishment for mutual defection). Note that the only requirement to 2The value of q for which E(C) = E(D) is q = (P - S)/(R - S - T + P). Thus, Sq/SR = -(P - S)/(R - S - T + P)2, which is negative for all P > S. 104 change the PD into an AG is that in the former T > R while in the latter R > T. In both cases, however, P > S . The only reason the value of P was changed from Figure 3.1 to Figure 3.2 was to illustrate the increased likelihood of punishment and/or the increased psychic distress as a result of noncooperation. The purely mathematical results are similar whether T is decreased or P decreased (increased in absolute terms). Reducing the first and increasing the second (in absolute terms) has made cooperation more likely. It is also conceivable that R and S may change due to social control. An example would be when the act of cooperating does not impart merely tangible benefits but also psychic satisfaction; in such cases, R would go up and thus render cooperation more likely. Similarly, S could increase (that is, go down in absolute terms). This could be the case, for example, when even though a person may not benefit directly from the group project, (s)he enjoys some satisfaction from the group's success. The assumption of P > S seems reasonable perhaps except in the case of masochists. In the general case, it is worthwhile for a risk-neutral person to cooperate iff: 105 q > (P-S)/ (R—S-T—P)3 I denote this q, qjj, that is, the minimum probability of j's cooperation that makes it worthwhile for i to cooperate; henceforth I shall refer to q.. as the probability threshold. 3.4 Interdependence and Cooperation Let us now consider the (interdependent) decision-making process facing i. The true qj, which is j's probability to cooperate, is not observable. i forms an estimate, p^, of qj, based on whatever information (s)he has about j, such as the latter's past behavior, preferably vis-a-vis i or with people i knows, j's preference function and so on. If Pj < q^-, i will not cooperate. But the result is not symmetric. If (s)he cooperates with certainty whenever Pj > q.., and if this information is available to j, the latter would defect and thus leave i with the sucker's payoff. This result follows because we are assuming self-interested, utility-maximizing individuals. What will i do? (S)he will cooperate with probability q4 if 3This can be derived by recognizing that E(C) > E(D) when Rq + S(1 - q) > Tq + P(1 - q). 106 pj > q^-. As is obvious, j cannot observe the true qi either. Thus, in an analogous way, (s)he will cooperate with probability qj iff his/her estimated pi > q^. Thus, we have: g* = qj(Pj) qj = qjtej) An equilibrium exists if: Pj = qj(P,-) p, = q,-(Pj) To the extent that expectations are not realized, costs are imposed on the party making the error. This is why cooperation may well be more difficult to achieve in the first round than in subsequent rounds. This important, perhaps even vital, point is largely overlooked in the literature. As the game is played over several rounds, it is more likely that, with the increased information available, the estimated p{ will be closer to the true qj and that expectations will be realized. It is also noteworthy that social norms, beliefs, and customs may be crucial if cooperation is to be attained in the first round. Such factors may be a 107 substitute for the costly information-gathering process of several rounds. If i and j share a strongly held social or religious injunction concerning the discharge of all outstanding debts, it is much easier for either to correctly estimate the other's true q. Thus, in many communities, much more cooperation is observed than would be expected in the absence of such social factors. A rotating-credit association could not succeed if a person, who receives an early payout, leaves the group. Not only must (s)he not leave the group but this information should be available, with a high degree of accuracy, to the other members of the group. It is obvious that it may be too costly to gather such information by playing the game several rounds. It is thus not surprising that such associations are typically formed among neighbors or fellow office-workers, that is, people about whom there is substantial mutual information. It is important that the information be mutual; i must be well-informed about j and so must j be about i. Otherwise, even if one player's expectations are not realized, cooperation becomes more unlikely in subsequent iterations. 108 It is clear that anything that, ceteris paribus. increases q is likely to enhance cooperation. Social norms, for example, will tend to increase q and physical proximity will tend to make it easier to obtain information about the actual q with greater accuracy. In this context, the impact of mobility— the focus of Chapter 5— appears crucial. The more mobile a society is, the more difficult it is, ceteris paribus, to obtain information about the other members of the group and thus the less likely will first-round cooperation be. What about an individual's private payoff? It is now time to relax the ceteris paribus assumption. 3.5 The Model The method of exposition chosen is to first set out each equation individually and to examine it rigorously. Each variable is explained and the restriction(s) on these variables, if any, is(are) specified. Finally, the entire model is summarized. 109 3.5.1 What Determines Control The first equation sets out the functional dependence of social control and may be written as: (1) C-j=Cjj (M, Cl, w, n) + + ? Following Hechathorn [1988], c.. is the proportional reduction in i's violation probability (Vj) attributable to j's control. If i and j are both individuals, c.. is an index of interpersonal power and for all i=j, cii is one's control over oneself, that is, internalized norms or beliefs. Such cases are represented by changes in one's subjective utility which is captured by B{ 4. Hechathorn [1988] does not discuss the second possibility nor, and perhaps more importantly, does he examine the causality of c... This is the purpose of the first equation and the restrictions on the values of the variables and their derivatives. Mobility (M) can be expected to have a negative effect on interpersonal control (cM < 0) . In an immobile society, groups may know each other for generations and it is likely that they share certain 4This is examined in Section 3.5.3. 110 values and customs. Furthermore, and perhaps more importantly, monitoring is easier and less costly. There may also be feelings of camaraderie and friendship. One would expect c^. to be higher, ceteris paribus, in rural Bangladesh than in downtown Los Angeles. Among all the determinants of c^., mobility may well be the most important. Consider a society which is very mobile, that is, one where many members frequently leave and many others frequently join. The new members are not aware of, and/or do not share in, the beliefs and norms of the group they have just joined. Time and effort would be required whereby the new entrants would either assimilate with the existing members or perhaps not do so completely. In the latter case, the norms would have undergone (some degree of) change. Thus, we observe that in traditional societies norms are passed on from generation to generation while in modern societies norms change more frequently. This is the impact of what Basu [1991] calls "common teaching": a father teaches his children and teaches them to teach their children, and so on. An example would be the Muslim belief of personal responsibility for all actions. It Ill is clear that such a belief is less likely to persist if children and parents do not live for long periods of time in the same locality. Closure (Cl) is used in the sense of Coleman [1988] discussed in Chapter 2. If it is possible to influence others' externalities, closure is said to exist. A simple diagram depicting relationships among people will help clarify this concept. D !B B C \ / \ / : \ / \ / \ / \ / \ / \ / \ / A Figure 3.3 Figure 3.4 Thus, closure exists in Figure 3.4 where, since B and C have relations with each other, they can combine to sanction A but does not exist in Figure 3.3 where B and C have no relations with each other but only with different third parties D and E respectively. Control is more likely when closure exists than when it is absent. An implication would be that persons linked in 112 a multiplex relationship (neighbor, fellow worker, coreligionist, parents whose children go to the same school) are more likely to control, and be controlled by, one another. How could one measure closure empirically? One method would be to consider the ratio of existing interpersonal relations to the total number possible (or important) within the relevant group. The more people in a group of a given size that cooperate, the more likely it is that others can influence each other; thus, c > 0. This follows ' W because the higher w (the proportion of the group that cooperates) is, the easier it would be to monitor and the higher would be the costs of defection. The well-known Japanese proverb "The nail that sticks out is flattened" is a succint description of this situation. Thus, it is observed that, in many Muslim societies, moneylenders who charge interest do so discreetly and/or under the veil of various ruses. The restrictions on n or the number of members in the group— that is cn" < 0, cn‘ > 0 when n < n* and cn' < 0 when n > n*— imply that c goes up with n until n* is reached, after which it declines. 113 Why is this the case? And what is the value of n*? If the group is too small (n < n*), the type of social give-and-take required to exercise control is unlikely to develop. The probability of a school student playing hookie getting censured by a parent may be increased when that student's parents are members of a close social group of four sets of parents (for example) than when they are members of a group of only two sets of parents. This is because more resources (that is, number of rules monitors and enforcers) are available for monitoring violation and enforcing sanctions. At some point (n > n*), the group may be too large in the sense that monitoring is more costly and difficult and the probability of punishment in the case of defection is too remote. If the number of parents is five hundred, for example, it is clear that, due to the increased likelihood of free riding on the part of the rules enforcers, violators are more likely to go undetected. How large is large? This is an empirical question and is a function of the context. In a credit group with collective sanctions, as in the Grameen Bank example to be discussed in Chapter 4, n=5 114 seems to be ideal. In rotating credit associations, larger groups (n=20) have been successful. In addition to monitoring costs, there is an informational problem. If the group operates as a peer group— a working group with income-sharing or -generating arrangement but with no hierarchical structure— it is expected to work like a democratic council where unanimity is required for all decisions.5 An all-channel network is essential in such cases: 5An example of this was seen in Davis, California. A new type of housing development was initiated to enhance community feelings. The innovations included garages at the back of the house rather than the front, common dining spaces where families cook by turns and common playgrounds for the children. Although the people enthusiastically endorse the experiment, it has taken them over eighteen months to agree on the fence to construct around the playground. A hierarchical structure— appointing a committee, for example— may well have made the decision-making procedure easier. Note that the information costs are particularly high for large groups. 115 1 2 3 4 Figure 3.5: All-Channel Network The number of channels, Ch, required is thus n(n-l)/2 where n is the number of people in the group.6 As n increases, decision-making becomes more costly because more time and resources are expended in soliciting and receiving members' views on different issues. At some point, for n > n*, the group becomes too large in the sense that it is more and more difficult to have unanimous agreement on any issue. This is possibly why democratic communes as distinct from those imposed by a centralized authority are not likely to be very large in size. c.j has been defined in the unit interval, that is, 0 < Cjj <1. In atomized groups, cf j . = 0 while in highly centralized groups, c,. = 1. Although not 6Figure 3.5 shows that if there are four people, there needs to be (4 x 3)/2 = 6 channels. 116 considered here, it may be interesting to imagine cases where -1 < Cjj < 0. These would be instances where, for example, the views, advice or recommendation of some people lead one to do the opposite. 3.5.2 Mean Control The second equation is a measure of the mean control of the group over i, thus: (2) C i =th^ t £ c« =Ci<“ ' c + 1- r - ? Vi*j,Vi,j=l,...,n The mathematical justification for defining the average group control in this manner is to ensure that 0 < Cf < 1 and thus that the units are compatible with those of the other variables. What is the intuitive reason? When there are several (or many) people in a group, each member j will usually have some control, cf J ., over i. This control tends to be averaged out. Thus, one may not be controlled at all (C{ = 0) if there are two persons in a potential group, one of whose views one does not care for at all (c^. = -1) and the other is a 117 person whose views or advice one accepts completely (c^ = 1). In this case, the mean control Ci = 0 and i does not feel controlled, on the average, by this potential group. As is obvious, intermediate values of Cjj between 0 and 1 are more likely to occur and can be easily averaged out. Both Cjj and C , . can, on theoretical grounds, assume negative values as discussed above. It is, however, quite likely that, in most empirical situations, C, will be nonnegative even though some c^- ' s may well be negative. This provides intuitive reasoning for using the assurance game because of what Platteau [1992] called the "predisposition toward cooperation". As noted above, what we have done is imparted valid theoretical reasoning for using the AG. Since our primary focus is on Cs, the simplifying assumptions of nonnegative c^. and C? are not restrictive. Indeed, the unit scale can be readily redefined to handle such situations, should they arise. For example, cij = 0 can be defined as an intense dislike of j by i (the equivalent of c^- = -1 above) but leaving the rest of the analysis unaffected. 118 3.5.3 The Private Utility Function The third equation states that the net addition to one's utility by cooperating (B^ is an increasing function of one's private payoffs, monetary and psychic. (3) Bf=B, (TM,, c. . ) + 4 - While evaluating his/her B,., the actor is not altruistic nor does (s)he endeavor to maximize social welfare. There are two separate influences on B{. The first, TMj, denotes the monetary (or tangible) component affecting one's utility. This could take the form of an increase in net expected income from undertaking some activity. In terms of microeconomic analysis, this is a change in the choice set and leads to purchases of more of all (normal) goods. The second, cH , however, denotes the internalized norms and beliefs, i.e., the extent to which one is controlled by oneself. This corresponds to a change in the arguments of the utility function itself. Consider a situation where a person believes that gambling is bad and should be banned. If such a ban should actually materialize, the result would increase 119 that person's private (psychic) well-being, even though his/her consumption of material goods and services would not have increased. This would be the case even if the person never gambled or even intended to gamble, i.e., even if his/her monetary (or expected monetary) loss from such an institution were zero. It is evident that there can well be instances where a person will not undertake an activity in spite of the fact that TMj > 0. This could be the case if one were to believe— or was convinced (by propaganda)- -that doing so would not be the right thing to do. This factor is in addition to considerations of long-term expected tangible (or monetary) benefits as opposed to those of immediate benefits. Thus, one may abstain from stealing because: (i) given the probability of getting caught, one's expected monetary benefits may be maximized by not stealing, or (ii) even if the expected monetary benefits are maximized by stealing (because, for example, the probability of being caught is small7) , one may refrain from such an activity due to religious beliefs. In this example, 7or because the expected return from stealing, if not caught, is very high or both 120 the effect of TMj on B, is positive, i.e., consideration of monetary payoffs alone would imply a net increase in utility by stealing. It must be remembered, however, that monetary or tangible payoffs are not the sole determinants of the net addition to utility, . The psychic or nonmonetary payoff, clM, must also be taken into account. Stealing will lower B{ if one's internalized beliefs (c^.) lead one to conclude that such an activity is bad. The net effect on Bi would thus depend on the relative magnitudes and signs of TM1 - and c-- It is not difficult to visualize situations where a person may not steal— even though it is monetarily rewarding— due to his/her internalized norms or beliefs. Since these effects are likely to be different for different individuals, it is not surprising that we observe some people who steal and others who do not. More importantly, this model would predict that some people steal because their expected monetary gain exceeds the psychic loss from doing something wrong. On the other hand, nonstealers may be of two types: (i) for some people (the amoral), the expected monetary gains are not large enough and (ii) 121 for others (the moral), the expected monetary gain is not sufficient to compensate them from undertaking such an undesirable activity. It is also worth recalling that most religions would not stop at merely stating that one should not steal because it is bad. There is a promise of an explicit (though not perhaps certain) promise of reward in the future, e.g., the attainment of paradise. It is thus clear that actions may be undertaken that apparently involve lower monetary payoffs than other alternatives.8 In some cases, even activities with negative payoffs may be carried out (if the psychic returns are high enough). An example would be a political activist who sacrifices (immediate) material goods for the sake of psychic returns (s)he gets for fighting for what (s)he perceives is a just cause. The cu term reflects precisely these types of private psychic gains. Bounds of family, kinship, friendship, and camaraderie are factors which are captured in c^. and 8Such behavior may be called "irrational" unless one considers psychic returns in addition to monetary returns. Thus, an incomplete specification of the utility function may lead one to call rational behavior irrational. 122 apparently irrational actors are, in reality, merely maximizing their own utility just like anyone else. If one believes that it is one's duty to care for an elderly parent, one may be willing to bear the costs of such an activity. It is important to note that we have not assumed altruism which may be defined as "the principle of living for the good of others". Actors are acting in ways that yield maximum satisfaction to themselves. Their actions may well do good to others but this must be considered an externality (i.e., an unintended consequence) rather than the primary motivating factor. This model would stress the fact that one would do good unto others primarily because it increases one's own satisfaction (or reduces one's psychic distress) and not because it is beneficial to others. The above arguments lead to a subtle, and often overlooked, distinction that is important in empirical situations. The educated elite in many Third World countries may be willing to work at home for a pittance even though many could be much better off financially by working abroad. They may perceive a need to stay home evolving out of kinship ties. The crucial point 123 is that they usually do not stay home out of a desire to render good unto others but because they enjoy psychic benefits themselves. Similarly, many well-to-do expatriates suffer emotional distress out of a sense of not belonging. It is noteworthy that, while the latter situation is well recognized, the former is not. Among the common motivations ascribed to highly qualified Third World technocrats choosing to stay home may be: (i) irrationality and (ii) an altruistic desire to be of service to one's country and its people. Either of these motives can be handled by the model being developed. In the first instance, the agent makes a suboptimal choice due, for example, to an information problem. The second case would be what I would term the "true altruist". For this person, an increase in someone else's utility does, in fact, raise his/her own. The c^ term would then show an increase when someone else's utility increases and this would imply, ceteris paribus, an increase in one's net addition to private utility, Bi. Casual empiricism suggests that the true altruists are much fewer in number than what might be called the "disguised altruists". In other words, such people choose to stay 124 home for a variety of self-interested reasons like an increase in self-worth out of a feeling of having properly discharged kinship ties or proximity to the levers of power which is easier at home than abroad. It thus seems reasonable to continue with our assumption of self-interested utility-maximizing individuals while recognizing that altruists may exist but that (a) they are likely to be relatively few in number and (b) their presence can be handled by the model with minor adjustments. Finally, it is important to reiterate that altruism is not necessary for group cooperation which is the result of self-interested individuals maximizing their utilities subject to whatever constraints that they may face. 3.5.4 Interdependence and Private Benefit Turning now to the fourth equation, qM is i's estimate of j's probability threshold. (4) qji=qj (Bf) for all i=j , i, j=l,2, . . . ,n In other words, in an interdependent situation, it is the minimum probability of j's cooperation that makes it worthwhile for the risk-neutral individual i to also cooperate. As shown above, this qM is the probability 125 threshold. The true probability of j's cooperation is, of course, not known to i. is the net addition to one's utility by cooperating and is defined, for ease of exposition, on the unit interval. This equation essentially states the higher is , the lower is q^.. This means that when i's expected utility gains are high, it is worthwhile for him/her to cooperate even if j is less likely to do so. 3.5.5 Mean Group Threshold The fifth equation defines a mean threshold where Qf is the group's mean threshold. (5) Qi= qji' Vi*j, Vi,j=l,...,n In a group setting, we consider the average thresholds of all (potential) members. This allows for the probability that, even if some members have high q^.'s, the countervailing effect of others' low qj/s may yield a relatively low average Qi. What does a low Q. mean? It means that it is more likely that i will cooperate. This may be particularly important when some members are not excludable, e.g., in kinship 126 groups. The specific mathematical form that takes ensures that 0 < Q- < 1. 3.5.6 Determinants of Defection Finally, we turn to the most important equation of all. The sixth equation specifies the functional dependence of Vf which is the probability that i will defect in the presence of control. (6) Vi=Vi (Dj, cf, Q(# Bf) + - + - Naturally, the higher is the value of Df— violation probability without control— the higher would be V,., ceteris paribus. But the impact of Ci is negative. What this means is that, for any given D{, control makes violation less likely than would otherwise be the case. In the atomized world, C, = 0 and Df = V. and people are not affected by the social context.9 Most of our day-to-day minor market transactions are of this type. In a totally regimented society, C , - = 1 and Vi = 0, regardless of the value of D^. The popularity of a war cannot necessarily be measured by the number of defectors. There may be no defectors because the 9The results for a linear model are shown on page 31. 127 soldiers genuinely believe that they are fighting for a worthy cause, that is, Df = 0. On the other hand, even a universally unpopular war with D? = 1 would have no defectors if Cj = 1, that is, if the commanders were to have total control over potential defectors. A large value of C- may be the result of an active policy of publicly executing deserters and confiscating their property. The positive sign of the relationship between Q. and V1 - is intuitively obvious. But what does a higher value of Qi mean? It means that the (average) minimum probability of others' cooperation that makes it worthwhile for i to cooperate has increased. In such a case, V1 - is likely to go up, that is, i is more prone to defect. V. depends negatively on B? because we have assumed a self-interested, utility-maximizing individual. Thus, if one's net addition to private utility— monetary or psychic— is increased, one is less likely to defect. It is to be noted that the decision to defect is not determined solely by tangible (or monetary) considerations but also by those pertaining to internalized norms and customs. This is precisely 128 why propaganda aimed at influencing one's beliefs (the cH 's in our model) are used so frequently. The formal model proposed above may be summarized as follows: (1) cij=clj(M, Cl, w, n) - + + ? (2) Ci = — 1 - Cl, w, n) (n-lj - + + ? Vi*j, Vi,j=l,...,n (3 ) Bi=B1 (TM±, cii) + + (4) qji=qj (B±) Vi*j, Vi, j=l, . . . ,n <5)0‘- <n-l>£^ Vi*j, Vi,j=l,...,n (6) V1=Vi(Dit Cit Qi, Bi) + - + - where the endogenous variables are cfj, C^, Bf, q.., Q , - ' and V.; the exogeneous variables are M, Cl, Df, and TMf 7 129 where w and n are the predetermined variables and where Cj. is control over i by j M is mobility Cl is closure w is the proportion of group cooperating n is the number of members in the group B| is the net addition to one's utility by cooperating TM,- is the monetary (or tangible) payoff from cooperating Cjj is the psychic (or nonmonetary) payoff from cooperating qjj is i's estimate of j's threshold Qj is the group's mean threshold Vi is probability that i will defect in the presence of control, that is, when 0 < Ci < 1 is probability that i will defect in the absence of control, that is, when Cf = 0 The restrictions on these variables are: 0 < Vi, Dj, Bj, qjf, Q,-, Cj j, Cj < 1 cM < 0; ccl > 0; cw > 0; CM < 0; Ccl > 0; CH > 0; Btm > 0; B,. > 0; c" < 0 and in c n 130 cn‘ > 0 when n < n* and cn" < 0 when n > n* 3.6 The Simplified Model The above six-equation model can be collapsed into a two-equation one by algebraic substitution, thus10: (A) c, = Cf (M, Cl, w, n) + + ? (B) V, = Vi (Ds, Cj, Qj, B1) + — — Endogenous variables: Ci, V,- Exogenous variables: M, Cl, D}, Qf, Bj Predetermined variables: n, w Note the recursive nature of the model. In (A), C. is determined as a function of M, Cl, w, and n. This value of C{ then enters (B) and, along with D,, Q,, and B., helps determine V^.1 1 Thus, Vi can be affected 10This can be done by the following substitutions: (1) into (2), (3) into (4), (4) into (5), and (5) into (6) . 1 1 As noted below, the value of w (and n) will, in general, change and thus, at the next iteration, w (and n) will enter (A) with a different value. This provides the mechanism in this model for control to be strengthened or weakened over time. 131 either directly through a change in D{, Q5, and Bf or indirectly through a change in C,. In any empirical situation, Df is not likely to change much unless the payoffs themselves change. Bj could change as shown in (3) due to a change in monetary and/or psychic returns. A change in B? would lead to a change in by (4) and that would lead to a change in Qi by (5) . For the same activity— a prevalent norm like caring for one's elderly parents, for example— a change in Bi is not likely to be of a huge magnitude unless the psychic component cn undergoes a radical transformation. Casual empiricism suggests that changes of this type are not so sudden. In the absence of a change in Bi, q^. will not change unless other people are somehow more or less likely to cooperate. The above discussion leads us primarily to look at (A) for changes in Vi. n and w are predetermined in this two-equation model; they each enter the model at any time period with a specific numerical value and, if there has been any net change in the number of cooperators, they each assume a different numerical value. There are then two variables, M and Cl, which would appear to drive this model. If mobility 132 increased, C, would decrease and that would lead to an increase in Vf. It also needs noting that, to a certain extent, even Cl would be increased by lower M. In a mobile group of people, the type of multiplex relationships that are required for closure are less likely to develop and coalesce. This does not mean that mobility is the only variable initiating changes in the model but that it is likely to be the one that causes the most sudden changes. Note that changes in c^., for example, will also affect V , - through its effect on Bf. 3.7 Statics and Dynamics The model presented above can explain the dynamics of norm formation and decay in addition to the static effects of control. At any point in time, tQ, w and n take on fixed values determined in the previous time period t.1. The control is then determined by these values of w and n in combination with M and Cl.12 This Ci, along with D5, Qf, and Bj, determines V1 - by (B) . This is the static picture of control in the model. 12See equation (A) . 133 The dynamics are embedded in the recursive nature of the model. At tQ, as noted above, w and n will usually be different from the values these variables took at t.,,. Thus, even if none of the other variables change, the realized value of Vi at time t1 will be different from that at time tQ. This could explain why some norms are fragile in that if some people defect it leads to a complete decay of the norm, even though other conditions have not changed.13 A further cause of dynamic change is B1 - which is a function of TMi and c^..1 4 If tangible benefits increase, the likelihood of defecting goes down and the norm is strengthened. Similarly, if psychic returns from an activity goes up, the norm is more likely to 13The flip side of this is that as a larger proportion of people support a norm, it gets solidified in society. The results would tend to support the views expressed in Kuran [1987]. He seeks to explain why institutions like the caste system persist and are defended even by victims like the lower caste people. He argues that punishments are meted out to opponents of the system and most opponents do not publicize their beliefs. New generations hear much in favor of the system and little or nothing against. In terms of our model, the cii term is affected by propaganda whereby individuals come to believe that the norm is beneficial, thus strenghtening its hold on society. uSee equation (4) above. 134 persist.15 Note that changes in Bi (as well in Df and Qf) causes changes in V1 -, independent of the changes caused by Ci. The system is thus not a static one but is best viewed as one of perpetual motion. 3.8 A Linear Example It is convenient at this stage to consider a simple linear form of the above model and attempt to trace out the comparative statics. Such a form could be: (E) Vi = D, [1 - C , . ] (F) Q. = V./tl - B,] Endogenous variables: C,., V. Exogenous variables: M, Cl, D1 -, Qf, B,-. Note the functional dependence of the variables has been dropped for notational ease and the predetermined variables of the above model— n and w— have not been incorporated for ease of exposition. As will be demonstrated below, none of this causes any loss of generality. 15This is perhaps what Basu [1991] calls "minimal functionalism". In terms of our model, a highly beneficial activity (that is, one with high Bj) is more likely to be engaged in. 135 A two-dimensional graph of this model is shown below: vi Vi slope of L=l/[1—Bio] Slope of M=l/[1-Bil] Vil Qi Q* slope of J=l-Cio Slope of K=l-Cil ViO vil Di D* Figure 3.5 Figure 3.6 Figure 3.5 is a depiction of (E); if control increases, goes up from Ci0 to Cn . This would be due to the effect of M, Cl, w or n as modelled above. The line pivots to the right because the slope decreases from (1 - Ci0) to (1 - Cn) . For a given violation 136 probability without control of D*, the violation probability with control decreases from Vl0 to Vn . This is the effect of the social context: history, culture, religion, norm. Figure 3.6 depicts (F); if one's net private utility goes up from Bi0 to Bn, the line again pivots to the right. For a given Q*, the violation probability with control falls from V., to Vi2. Cooperation is thus more likely because the ith person is less likely to defect. Note that the final level of V. is the result of both the social context and private benefit. In this example, both enhance the likelihood of cooperation and work in the same direction. It is easy to visualize cases where the private benefit may fall and would tend to offset the effects of increased control. similarly, decreased control may interact with increased private benefit. The linear model can be simplified further in order to examine carefully some other features of this model. Simple algebraic manipulation yields: (1) V, = [ 1-B, ] [ 1“C- ] Q( Dj 137 When Bf = 1, that is, the private benefit is extremely desirable, Vs = 0 which means that i will cooperate with certainty regardless of what others do. When Bi = 0, that is, the private payoff is nonexistent, is, in general, not equal to one primarily due to the effect of C{. Thus, cooperation is enhanced by social control even when it is not judged worthwhile at all by purely private utility considerations. When C< = 1, V. = 0: this is the case of the unpopular war referred to earlier. Even if Di = 0— there is universal hatred of the war— there are no defectors. When Cj = 0, we obtain the atomized solution. What can we say about the dynamics of norm formation and decay? Perhaps the crucial role is played by w, the proportion of a group cooperating. At any point in time, w is a predetermined variable, that is, it enters into the period with a certain value. But, at the end of the period, it will usually have a different value. This will be the case whenever the net number of cooperators (i.e., new cooperators minus new defectors) is different from zero. And this leads to different initial conditions in the next period. In 138 threshold models, it is shown that, whenever preferences are interdependent, small differences in the distribution of disposition can lead to big differences in group behavior. This would imply that C- is very sensitive to w. Note the logical justification for denoting w as a predetermined variable; the recursive process changes the value of w and that in turn affects Ci. It is also obvious why control could be very fragile as a means of ensuring cooperation when M increases or Cl decreases. Not only is there the direct effect of the variable concerned but there is the additional effect of w. In such situations, beneficial norms need to be backed up by increasing net private benefits accruing from the initiation or continuation of a certain activity, in the form of subsidies, for example. 3.9 Whv the Model Can Explain Failure and Success The model developed above is a general one and can explain both success and failure in group cooperation. It pays careful attention both to the micro-level utility-maximizing decisions and to the (usually) 139 macro-level social constraints.16 I prefer to call an internalized norm or belief a micro-level effect on the utility function rather than a constraint. Unlike much sociological literature in the methodological holism mold, failure to cooperate need not be regarded as ' ’deviant" behavior.17 At the same time, it is explicitly recognized that individuals are usually not completely atomized. Thus, an atomistic model may conclude that people are behaving "irrationally" whereas the problem is due to incomplete specification: the overlooking of norms, for example. In terms of our model, this would be overlooking the cn component of the utility function in (3). 3.9.1 Methodological Individualism Decisions about the course of action to undertake (cooperate or defect) are taken on the basis of calculations with respect to the maximization of (expected) utility subject to various constraints. Since we are assuming multi-period games, short-term benefits may be less important than long-term ones. 16Note, as mentioned earlier, these constraints may or may not be conducive to social welfare. 17See the stealing example above. 140 Suppose there is a norm which states that members of a village should work within that village. It is believed that migration to other villages or, worse still, to towns would corrupt villagers. Methodological holism would conclude that all villagers should (and would) work locally and that anyone who did not do so was a deviant, perhaps to be shunned by others. Methodological individualism would attempt to measure the net benefits of migrating and would conclude that a person who does not migrate when his net benefits are positive is behaving irrationally. In some instances, "psychic costs" may be mentioned almost as an afterthought. Let us consider a young man who finds a high salaried job in the capital, such as to assure that the net monetary benefits are positive. In the present framework, he will balance these monetary benefits with the costs, if any, resulting from control. In the extreme case where C , . = * 1, V. = 0 even if B , - = 0 (Bf is being considered for working within the village). As control is relaxed, V; would move away from zero and people would be more likely to migrate. It is to be noted that if this norm is internalized, staying in the 141 village would give the person satisfaction in and of itself. In such a case, he might be willing to sacrifice the higher monetary returns for his psychic pleasure. Can we label such behavior irrational? It appears we cannot unless we analyze much more rigorously the domain of preference formation. It is clear that, if the norm is not internalized and if Ci is not equal to 1, more and more people will start migrating. For each person i, there is usually a different threshold at which he migrates. Some people may be willing to dodge bullets while others are only willing to take a comfortable plane out and the majority may be in the middle of these two extremes. But, as more and more people start migrating, the control mechanism itself decays (as w goes down, C , - goes down). This is an example of why norms can be effective while they last but can also be extremely fragile as a control mechanism. People must be receiving benefits either in the form of monetary payoffs or in the form of psychic betterment. It is easy to understand why totalitarian governments prefer propaganda over coercion unless their rule is imminently threatened. 142 3.9.2 Hypocrisy and Cooperation Hechathorn [1988] investigates the role of collective sanctions where an individual's violation of, or compliance with, a rule or norm leads to collective punishment or reward enforced by an external agent. In Hechathorn [1989], it is shown that the sanctioning syatem is itself a public good since its benefits accrue to all members of the group. Thus, there arises the second order free rider problem of the maintenance of the sanctioning system itself. There are four possibilities: (i) full cooperation; (ii) private cooperation; (iii) hypocritical cooperation; and (iv) full defection.18 He finds that second-order cooperation is more robust than first-order cooperation but expresses surprise at his findings. In terms of our model, such a result would be entirely expected. The costs would be lowest for a hypocritical person: such a person does not bear the costs of sanctions but nevertheless enjoys the public good. The of the hypocrite would be relatively high since as long as the control is not 18For details, see 2.2.7 above. 143 1 binding, TMi would be high. The only offset would be the c^- term, that is, the extent to which his/her internalized beliefs are affected. For a true hypocrite, the effect on the c , . , . term would likely be minimal. Recent examples include the call of Saddam Hussein (an avowed secular socialist) for an Islamic holy war and the activities of TV evangelists like Jim Bakker. It appears that the reason Hechathorn is surprised is that he does not formalize a private payoff function and is thus unable to consider micro-level behavior explicitly. Such an examination would be unlikely to reveal Hussein and Bakker to be truly religious people. 3.9.3 The Importance of Information The above model also helps explain why cooperation may be attained in some situations while not in other very similar ones. An important role is played by the information transmission mechanism. We reject the functionalist argument that an institution will necessarily evolve merely because it is of benefit to society and that it will wither away when it is harmful. Recalling that we do not know the actual probabilities of others* cooperating, we are guided by 144 estimates of these on the basis of the available information set. The same is true for every other member of the (potential) group. It is thus clear that— to the extent that these estimates may be incorrect— cooperation may or may not result. The model is set out in probabilistic terms: the presence of certain factors will facilitate, though not necessarily ensure, cooperation. It would not be surprising to observe different outcomes if there are different informational conditions. The link to Olson's [1965] political entrepreneur is obvious. Such a person would be one who (a) had access to more information, and (b) would be able and willing to disseminate it, thus rendering cooperation more likely. Why would (s)he undertake such a function? (S)he could gain materially, or in terms of increased power and fame, or (s)he could experience psychic satisfaction at having done something worthy. In Chapter 4, it will be shown that Yunus performed such a role in the Grameen Bank case. It is important to understand why people would cooperate in the first round. In general, transactions involve reputation, trust and rapport. People who 145 interact have history, culture, norms. The greater the number of common elements among them, the more likely it is that cooperation will ensue. This is because each member of a potential group will be better able to estimate the others' likely behavior more correctly than when strangers attempt to form groups. In the latter case, the role of political entrepreneurs may be pivotal. Since first round cooperation is crucial, this model would predict that voluntary cooperation is more likely among homogeneous people. What can we say about subsequent rounds? Would someone buy an orange from a store that sold rotten oranges last week? The answer would be that if transaction costs are not too high (if the alternative grocery store is five miles away, Bi would be lowered considerably by going there), a rational person would consider shopping elsewhere. Thus, even in a situation expected to be atomistic, reputational factors are important. One has to remember that, when we observe cooperation or defection, we are looking at a slice of time. What happens now is affected by what happened in the past and will affect what happens in the future. The model developed accomplishes precisely 146 this dual objective: it helps us understand the effect of control on behavior at one point in time and the mechanism through which the controls themselves change over time. Methodological individualism was assumed throughout and careful attention was paid lest the theory be reduced to functionalism. Thus, as noted above, the model can explain both success and failure in group cooperation. Individual utility-maximization constrained by, inter alia, social controls leads to the observed results. The individual is not merely statically rational; (s)he is dynamically rational. (S)he forms expectations about his/her utility from different actions and undertakes the one that maximizes his utility over time. The social context evolves out of the activities of such cold, calculating actors. It was also demonstrated that altruism— at least in the sense in which the word is commonly used— is neither necessary nor sufficient for cooperation. The model's behavioral implications would tend to support Granovettor1s [1985] observation that the social context "not only shapes its members but is also shaped 147 by them, in part for their own strategic reason.”19 3.9.4 Generality of the Model As noted in Chapter 1, there are two main intellectual streams which seek to explain social behavior: the economic and the sociological. There has not, however, been an attempt to rigorously formalize a general model combining features of both. On the contrary, there has been substantial sniping at each other. This is exemplified by Duesenberry's [1960] quip that "economics is all about how people make choices; sociology is all about how they don't have any choices to make." There has not been any appreciable attempt by economists to study and understand social control apart from grudging remarks that such things exist. Although sociologists like Hechter, Granovetter, Coleman, and Heckathorn adopted methodological individualism, it is unfortunate that there were not many economists who endeavored to study sociological processes. One important reason for this neglect of social relations is offered by Granovetter [1985]. In 19emphasis added 148 classical and neoclassical economics, actors having social relations with one another may be a frictional drag, impeding competitive markets. Smith thus complained that "people of the same trade seldom meet together, even for merriment and diversion, but the conversation ends in a conspiracy against the public, or in some contrivance to raise prices."20 He suggested repeal of regulations requiring people in the same trade to sign a public register. This would minimize the probability that people of the same trade could meet and fix prices. Social atomization was thought to be a prerequisite to perfect competition. Contemporary work by economists either disregard the social influences entirely or refer to them almost as afterthoughts. Social influences are construed as processes "in which actors acquire customs, habits or norms that are followed mechanically and automatically, irrespective of their bearing on rational choice."21 This is reflected both in Duesenberry's quip above and in Phelps Brown's [1977] disparaging description of the 20Smith, Adam. (1776) 1979. The Wealth of Nations. Edited by Andrew Skinner. Penguin, Baltimore. 21 Granovetter [1985]. 149 "sociologists' approach to pay determination" as based upon the assumption that actors act in "certain ways because to do so is customary, or an obligation, or the "natural thing to do," or right and proper, or just and fair." There is no attempt to understand why social norms or controls develop and how they affect human behavior. The model developed in this chapter is more general than either of the two views examined above. While the actor is a self-interested, utility-maximizing individual, (s)he is subject to constraints, including social control. But, unlike Duesenberry and Phelps Brown, social control does not determine the actor's behavior; it merely influences it. As in all standard microeconomic analyses, the actor compares (marginal) costs and benefits from his actions. We have defined mean group control C5 over the unit interval; except in the degenerate (and unlikely) case where Cf = 1, the actor is not a robot. (S)he may or may not cooperate in the group activity, depending on his/her (subjective) preferences. There is, an explicit private benefit function, Bj = Bi (TMf, cH). As set forth in Equation 6 above, the 150 higher is B1 -, the lower is the violation probability, Vi. An inverse relationship also holds between mean group control and Vi. An important feature of the private benefit function is that it explicitly considers two separate arguments: tangible or monetary payoffs, TMj, and internalized norms, cn . These elements enable us to analyze a wider variety of human bahavior than if we considered either control or private benefits alone. Let us assume that there is substantial control. This will, ceteris paribus, lead to an increased probability of person i's cooperating. In a typical PD (Figure 3.1) or AG (Figure 3.2) situation, such cooperation will also lead to an increase in i's private benefit. However, remembering our assumption of self-interested individuals, we did not conclude that i would cooperate with certainty. As noted above, this may induce j to defect and thus leave i with the sucker's payoff. The model merely concludes that i would be more likely to cooperate. What if the increased control leads to decreased private benefits? Then, by Equation 6, Ci would tend to decrease V. but would tend to increase V{. The net effect on Vi would 151 depend on the relative strengths of Ci and B1. Thus, social influences are not followed mechanically and automatically and private benefits are explicitly considered. It is important to remember that the benefits themselves have two components. An actor may undertake an activity which involves monetary loss, provided there is (sufficient) compensatory psychic satisfaction. Thus, some Muslims may forgo interest income because, to do otherwise, may be considered morally wrong and may result in psychic distress. It is to be noted that an assumption being made in this particular example is that they are not sacrificing interest income in the hope of attaining paradise (or avoid being punished in hell). If the latter were the case, the analysis need only consider tangible benefits, current and future. It seems reasonable to assume, however, that some actions of individuals are motivated by intangible gains, such as, sense of duty and access to power. There is no altruism involved here: the person is merely trying to maximize his/her own gains and is not necessarily considering others' gains. 152 This model enables us to study a diverse variety of situations, depending on the actual condition at hand. These situations can vary from atomistic ones (Ci = 0) to completely controlled ones (Cr = 1) and, more importantly, to the intermediate ones where 0 < C , . < 1. If we wanted to study the market for oranges, assuming an atomistic model would seem reasonable. If we wanted to study why Saddam Hussein appeared to be so popular in Irag even in ignominious defeat, we would have to recognize the importance of control in totalitarian societies. And, if we wanted to understand the success of rotating-credit associations, we would have to consider the effects of mobility and closure on control. 3.10 The Need for Caution The above arguments point to the need for a careful study of the causes of success of institutions. While institutions may persist even when they are harmful (due to, for example, information problems), they are less likely to do so than institutions which render benefits. A similar scenario arises for the initiation of new institutions. It cannot be 153 hypothesized that new institutions will evolve whenever they are desirable. All that can be said is that, if new institutions are beneficial, they are more likely to evolve. Moreover, institutions based on control (and not on private benefits to members) are likely to be fragile. This is due to the effects of Ci and B , . on _V- discussed above. Thus, the fact that the so-called communist parties in Eastern Europe fell like a house of cards is not surprising. The next two chapters will examine two institutions: the Grameen Bank in Bangladesh and old-age security schemes in India. As will be examined in more detail later, we should be extremely careful about the replicability of institutions. A credit scheme like the Grameen Bank may have succeeded in Bangladesh precisely because it was designed to fit the social fabric of that society. Trying the same experiment in downtown Manhattan may be much more difficult to implement.22 Only a proper understanding 22See Chapter 6, however, for an example in downtown Chicago where a group credit scheme like the GB has been tremendously successful. 154 of the social factors as explained in the model can prevent us from making colossal mistakes such as this. 155 BIBLIOGRAPHY Axelrod, Robert. 1984. The Evolution of Cooperation. Basic Books Inc., New York. Basu, Kaushik. 1991. "Civil Norms and Evolution", unpublished paper. Coleman, James S. 1988. "Social Capital in the Creation of Human Capital", American Journal of Sociology 94 Supplement: S95-120. Duesenberry, James. 1960. Comment on "An Economic Analysis of Fertility", in Demographic and Economic Change in Developed Countries, edited by the Universities-National Bureau Committee for Economic Research. Princeton University Press, Princeton, N. J. Granovetter, Mark. 1985. "Economic Action and Social Structure", American Journal of Sociology 91: 481-510. Heckathorn, Douglas D. 1988. "Collective Sanctions and the Creation of Prisoner's Dilemma Norms", American Journal of Sociology 94: 535- 62. -----------------. 1989. "Collective Action and the Second Order Free Rider Problem", Rationality and Choice, Vol. 1, No. 1, pp 78-100. Kuran, Timur. 1987. "Preference Falsification, Policy Continuity and Collective Conservatism", The Economic Journal, 97: 642-665. Olson, Mancur. 1971. The Logic of Collective Action. Harvard University Press, Cambridge, Mass. Phelps Brown, Ernest Henry. 1977. The Inequality of Pay. University of California Press, Berkeley. Platteau, J. P. 1992. Chapter 3 of unpublished book. 156 Rapoport, Anatol. 1983. Mathematical Models in the Social and Behavioral Sciences. Wiley, New York. 157 CHAPTER 4 THE GRAMEEN BANK IN BANGLADESH: WHY GROUPS SUCCEED 4.1 Rationale for Subsidized Rural Credit The previous chapter set out a model of group behavior; it attempted to identify, inter alia, the conditions under which groups are more likely to succeed in cooperative behavior. Among the implications of the model were: (i) groups are more likely to succeed when the private benefits— both monetary and psychic— to their members are higher; (ii) groups are more likely to succeed when social control is greater; and (iii) access to information plays a vital role in group formation. It is the purpose of this chapter to consider the Grameen Bank (GB) as a case study in order to demonstrate how important it is to pay close attention to social factors in the design of economic institutions. During the 1950's, low growth rates in less developed countries (LDCs) were attributed to the lack of capital and vicious circles of poverty wherein low 158 income levels implied low levels of savings, which in turn implied low productivity. This, coupled with the belief that the production function in the "modern" sector— mines, plantations, large-scale industry— was characterized by fixed technical coefficients, meant that increasing capital formation was both necessary and sufficient for economic growth.1 Viewed in this light, augmenting the supply of credit was the vital policy imperative. Commercial banks have proved inadequate with respect to mobilizing rural savings or providing rural credit. Procedures for loan approvals, including complex application forms, requirements of collateral and of the existence of prior bank accounts, have been disincentives for the rural people. Since the loan amounts tend to be small, the per unit cost of processing these loans are high. Deposit procedures are also cumbersome, especially if we remember that many rural people are illiterate. Finally, there is ^his assumption was embodied in the L-shaped isoquants where more output could be obtained, that is, the firm could move to a northeasterly isoquant only by an increase in the fixed proportion of both labor and capital. The two inputs are thus not substitutable. 159 information asymmetry: the (usually) urban-bred branch managers have little knowledge about their rural cousins. And the cultural distance between the urban and rural areas of LDCs augments this distance. As a result, governments have increasingly turned to specialized farm credit institutions to provide subsidized loans for agricultural uses. Von Pischke [1980] cites four theoretical reasons for the creation of such institutions: (i) the perception that farmers are poor; (ii) the belief that agricultural development cannot occur without credit; (iii) the axiom that the government should promote rural development; and (iv) the belief that supply-leading finance can stimulate agricultural development and productivity. Thus, as Gonzalez-Vega [1983] observes, "... credit problem was perceived as structural and the solution as institutional: to create special agencies to provide agricultural credit." The prevalence of ceilings on interest rates— the price of loanable funds— requires examination. Interest rates in almost all LDCs have been kept at artificially low levels; the real rates have often been negative. If capital is the scarce factor, then the 160 price of borrowing capital should not be negative. These low (and often, negative) rates mean that firms that have access to credit would invest in relatively more capital-intensive technology— exactly the opposite of what relative factor endowments would require. The theoretical underpinning for this policy can possibly be traced, as Fry [1988] argues, to Keynes' concern about the liquidity trap. He argued that there is a secular trend for the real interest rate to rise above its full-employment equilibrium level, and thus the incentive to invest is threatened. If the price level is fixed, and expectations about the future price level is static, reducing interest rates and satisfying an increase in liquidity preference can be simultaneously achieved. An interest rate ceiling at low levels can stimulate investment if accomodative monetary policy is pursued. The prevalence of low interest rates can also be attributed to political and religious objections to high or even nonzero interest rates.2 Deficit 2One only needs to remember that in many Muslim countries the taking and paying of interest is frowned upon, if not banned outright. In the Western world, the misgivings about usury are obvious from the myriad 161 financing provides yet another motive: government borrowing can be accomplished at lower cost. The private sector can be encouraged to undertake "desirable" investment which may be unprofitable at market rates. Next, there is the neostructuralists * argument that rising interest rates cause inflation through a cost-push effect, at least in the short run. Finally, there is the political economy aspect of credit which is often overlooked: rationed credit at subsidized rates is a convenient means of dispensing favors to the relevant groups and lobbies. 4•2 Myth and Reality As Shaw [1973] observes, these distortions of financial prices reduce "the real rate of growth and the real size of the financial system relative to nonfinancial magnitudes." Insufficient finance has led to financial repression. Only those institutions from which the government can expropriate seigniorage are favored. The formal banking system is encouraged because reserve requirements and obligatory holdings of anti-usury laws on the books. 162 low-interest government bonds are a source of public savings. By the same token, private bond and equity markets are discouraged because they would compete for funds with the public sector. Furthermore, low bank interest rates imply that it is more profitable for firms to borrow from banks— if they have access to these loans— rather than issue bonds at higher interest rates. One factor largely overlooked is that, while low interest rates may stimulate investment, by the same token, they would act as a disincentive to saving. Current consumption would tend to be favored relative to future consumption and, in the absence of compulsion, would lead to lower savings and thus reduced funds available for investment. The result has been that rural credit has largely bypassed the poor farmers for whom it was originally said to have been intended. Gonzalez-Vega [1983] estimates that only 5% and 15% of farmers in Africa and Asia and Latin America respectively had access to formal credit. There has, furthermore, been a concentration of these loans in the hands of a few large farmers. In a typical country, 5% of the farmers 163 received 80% of the loans disbursed. Indeed, less than 1% of the fanners received 80% of the subsidized credit, about 15% benefitted from the remaining 20% and about 84% have not shared in this process at all. This failure, it will be argued, is due to improper targeting and insufficient attention paid to monitoring and incentive mechanisms. The problems associated with commercial banks and specialized agricultural banks are summarized by Von Pischke et al [1983] as "limited access to, and high cost of services to farmers and other rural people, absence of savings facilities, undermining of repayment habits, and a lack of cultural compatibility of banks and customers." Credit policy has continued to consist, in the main, of providing subsidized credit to the fortunate few. There has been little effort at mobilizing rural savings, except those of the rural rich and the traders. The perception is that most rural people are too poor to save; and that those who are not spend on conspicuous consumption. 164 4.3 Do the Rural People Save: Theory Is the above stereotype correct? Saving is typically thought of as the acquisition of currency or liquid assets such as bank deposits. Borrowing, on the other hand, is viewed as a contractual arrangement whereby the creditor lends funds on the borrower's promise to repay, often backed up by some collateral. But, as Lee [1983] argues, these concepts are too narrow. Saving occurs whenever any current income or wealth is kept aside for future consumption; conversely, borrowing entails current expenses being paid out of future (expected) income. Accumulated savings would be net worth, that is, financial assets, land, livestock, equipment minus outstanding debts. There are several potentially important nonconventional forms of assets which are usually overlooked perhaps because they are more difficult to quantify.3 These include: (i) labor owed by a neighbor in return for past favors provided and (ii) the obligation of children to care for their old parents. In both these cases, consumption (of leisure 3This, however, does not necessarily mean that they are any less important. 165 or goods) were deferred, too. In the notation of the model developed in Chapter 3, the first may be captured by the control variable c-- in that person i owes an obligation to j and thus i's behavior is affected by this control. The second reguires examination of internalized norms denoted by cn ; if children believe that they should care for elderly parents, then such behavior on the part of the former constitutes a form of saving for the latter. Thus, whenever the receipt of income and the act of spending are not simultaneous, saving must occur. Furness [1972] lists four reasons for this non-simultaneity: (i) it is convenient to make periodic payments for a continuous service, e.g., monthly electricity bills; (ii) commodities with indivisibilities require, for their purchase (either in full or at least a down payment), buildup of wealth or borrowing, e.g., buying a cow or tractor; (iii) income flows vary over the life cycle; and (iv) the timing of production may not coincide with that of consumption, e.g., harvests occur 1-3 times a year but consumption needs are rather continuous. 166 As is obvious from these examples, saving— abstention from current consumption— is a common feature of rural life. In many cases, instead of being a "residual", saving precedes expenditure. A farmer sells his/her crop but does not buy clothes until a week or more later. It is not clear, and not important for our purposes, as to how long funds must be held before they can be considered savings. Suffice it to say that any income not spent instantly is saved. Another important consideration is that the savings threshold is likely to be lower in the rural areas of LDCs than in developed countries (DCs). Reasons include climatic conditions, dietary preferences, thrift habits, availability of fewer consumption goods, and lower demonstration effects. Thus, the stereotype that the rural people save less does not appear correct on theoretical grounds. It is important at this stage to distinguish between mobilized and non-mobilized savings. The former would include bank deposits and cash holdings and would be synonymous with the conventional operational definition of savings. The latter would include hoarded cash, the neighbor's labor, children's 167 obligation to parents, and purchase of land and jewelry. It is the latter component of savings which does not show up in official statistics. One may conclude that hoarding cash is irrational because it is more exposed to risk of loss through theft and it does not earn any interest. Why does the person not deposit the money at a bank? In addition to the problems of locating a bank nearby, the transactions costs would include filling out complicated forms by the (usually illiterate) depositor, waiting in line, and perhaps not being treated too well by the teller. The interest income foregone and the expected loss from theft may be less than the transactions costs involved with the formal banking system.4 The rural people, just like people everywhere else, make rational decisions given the constraints they face. In the absence of health and unemployment schemes, jewelry hoarding is understandable. Norms with respect to caring for old parents appear logical when one remembers that in most 4We are assuming that the saver does not suffer psychic distress (due to religious beliefs, for example) from accepting interest. 168 LDCs there are no pension or retirement schemes except in the formal sector. It seems safe to conclude that the rural people are not insensitive to incentives nor do they behave irrationally. 4.4 Do the Rural People Save: Evidence What is the empirical evidence with respect to rural savings? Because of the urban biases referred to above, empirical work on rural savings has been scant until recently. Empirical investigations tend to support the theoretical contention that the rural poor can and do save. Ong et al. [1976], in a survey of rural Taiwan, find the average propensity to save (APS) to have been in the 0.20 to 0.25 range in the 1960's. The marginal propensity to save (MPS) during this period ranged from 0.3 3 to 0.66. Saving was defined in this survey as total net household income minus total household consumption. Adams [1978] finds the following: (i) For rural Taiwan, the APS ranged from 0.19 to 0.31 for the years 1960-74. 169 (ii) For rural Japan, the APS increased from 0.10 to 0.22 between the years 1950 to 1973. (iii) For rural South Korea, the APS increased from 0.04 in 1965 to 0.33 in 1974, with the APS among small farm households being quite high (0.22 in 1974, for example). (iv) A cross-sectional study of rural West Malayasia suggests significant savings capacity. (v) A study of rural Punjab, India suggests that, during the 1960's, the APS ranged from 0.12 to 0.37. In some years, the APS of the smallest farm households averaged 0.30. What is obvious from these studies is that rural people can and frequently do save. What is perhaps even more significant is that households save more when they had profitable investment opportunities as they did in the cases of Taiwan, Japan and Korea, where increased incentives led to increased savings. A probably typical LDC scenario is etched out by Panikar [1961] who cites the results of a survey in Hyderabad, India in 1949-50. 25% of savings was used to repay debts, 48% for consumer durables and ceremonies, 9% for land purchase, and 18% for agricultural production. At 170 best, only about a quarter (9% + 18% = 27%) of the savings was being invested productively. Untapped rural savings in piggy banks, inside bamboo posts and under mattresses; capital tied up in real estate speculation; and the unemployed labor force: all these resources need to be freed up for productive enterprise. 4.5 The Grameen Bank in Bangladesh The situation is perhaps nowhere as desperate as in rural Bangladesh where even official estimates put the percentage of landless in rural areas at over 50% and the average size of cultivable land at 2 acres [Hossain 198 6].5 There is thus a limited prospect for wage employment mainly because of the availability of underemployed family labor. If the family desired to undertake an investment and required a bank loan, in addition to the general problems associated with rural banking mentioned above, there is the dire shortage of tangible collateral to offer as security to the bank. Unofficial estimates of landlessness go all the way to 8 0%. 171 It is this context that the Graxneen ("Village") Bank (GB) came into existence. Its creator Professor Yunus of Chittagong University believed that the poor could generate productive self-employment without outside assistance if they had access to credit on reasonable terms. He designed the GB as an organization which would "provide a reasonably dependable forum through which the banking system can extend credit to the landless without collateral."6 A pilot project was started near Chittagong University— where Yunus taught in the Economics Department— in 1976. The project was formally launched in November 1979 in two districts and in 1982 was extended to three others with assistance from the International Fund for Agricultural Development (IFAD). The government turned it into a specialized bank in September 1983 and it became considerably more spread. 6Yunus, M. 1982. "Experiences in Organizing Grass Root Initiatives and Mobilizing Peoples' Participation: The Case of Grameen Bank Project in Bangladesh", paper presented at the 25th Anniversary Conference of SID, Baltimore. 172 The GB is a self-selective group credit scheme. The eligibility conditions for membership are quite different from what one would expect them to be in the case of loan applicants of the formal banking sector. Only persons belonging to a household owning no more than 0.5 acres of cultivable land or assets equal in value to no more than one acre of medium quality land are eligible for loans. It thus targets that segment of society that is most unlikely to be deemed creditworthy by other lending institutions: the landless. In the formal banking sector, the provision of collateral is usually considered an important signal of creditworthiness. But in the GB framework, more importance is imparted to intangible signals, as will be examined later. At this stage, it should be pointed out that most members of GB would not meet the requirements of loan receipients at banks. We should also note, as Varian [1990] observes, that the goal of GB is not to maximize profits but "to encourage rural development". It is, however, important to study whether such institutions are basketcases or are self-sustaining. The repayment rate is perhaps the most important indicator of a bank's 173 self-sustainability: as will be discussed later, the GB is not a welfare program like many agricultural banks in LDCs but is a tremendously successful and profitable institution. 4.5.1 Structure of GB A bank unit— covering 15 to 22 villages— consists of a field mamager and several bank workers. They go to villagers' homes and explain the goals of GB to prospective members. To obtain credit, a group of five likeminded people must be formed. Each such group elects a chairman and a secretary and holds weekly meetings. There is a one-month period during which the members of the group must satisfy the GB staff of their integrity and seriousness, their understanding of GB philosophy, and their ability to sign their own names. After this "test", two members are given loans, repayable in weekly installments of 2% of the loan amount. Only when the repayment behavior of the first two loan receipients is satisfactory over a period of 1-2 months do the others become eligible for loans. Furthermore, collective sanctions are applied: if any one member defaults, the entire group becomes ineligible for loans. 174 Several groups are federated into a center and a center chief and a deputy center chief are elected from the group chairmen. These officials recommend loan proposals and assist GB staff in their work. Unlike other banks, loan recipients (or members) have direct operational control of the GB's lending policy. Only rarely are loan proposals recommended by the center chief or deputy chief turned down by GB officials. This helps solve the information problem inherent in a bank's lending decisions. As mentioned above, managers of commercial banks in LDCs have little knowledge of the creditworthiness of the rural people. The GB method means that such evaluations are being made by people who have the most information: that is, a peer group. Members are required to save Taka 1 (Taka 40 = 1 US dollar) per week and 5% of the loan amount. These savings form the Group Fund (GF) which caters to pressing non-productive needs of members. It is a common feature of many LDC credit schemes that recipients divert funds from production needs to consumption needs like medical expenses and children's marriages. Once this is done, the project for which 175 the loan was originally taken becomes infeasible and the person ends up being a defaulter. It is not enough to merely request or even threaten loanees not to indulge in such behavior. There may well be a rational reason behind such diversion. What would a typical father do if his child was dying and if he had no money except the loan he just obtained for buying a rickshaw? The GF helps ensure that such diversions will not occur. If a member quits the group, (s)he forfeits her/his claim to share in this fund. This not only facilitates group savings but also increases the exit costs of members and thus increases the stability of groups. In terms of our model, this means that, ceteris paribus, the private benefit of staying in the group, Bi, is higher than it would be if there were no exit costs. There is also an Emergency Fund (EF) that provides insurance against default, death and disability. Since there are collective sanctions, the importance of the EF to villagers living at subsistence levels can hardly be overemphasized. From its humble official beginnings in September 1983, the GB had grown to 961 branches with a 176 membership of 1,198,849 by May 19927. Hossain [1986] finds that the target group was almost invariably reached with non-target membership less than 5% and that only an insignificant proportion (<5%) of loans is diverted to unproductive uses. The contrast with the usual scenario in LDCs where loan recipients are the rich, the powerful or the politically well-connected is as stark as is that with respect to the productive use to which the loan is being put. 4.5.2 The Repayment Rate of GB The most remarkable aspect is the repayment rate: almost all GB loans are repayed within one or two years (see Table 4.1). 7These figures were obtained from: Grameen Dialogue. July 1992. Newsletter No. 11, Grameen Trust, Dhaka, Bangladesh. 177 j Table 4.1: Rate of Recovery of GB Loans, 1983-86.® Area of Percent of loans Percent of loans operation recovered within recovered within one year of two years of disbursement disbursement 83 84 85 86 83 84 85 86 Chittagong 99 100 98 98 100 100 100 100 Tangail 99 99 91 95 100 99 96 95 Rangpur 100 100 100 98 100 100 100 100 Dhaka 100 100 98 96 100 100 100 100 Patuakhali 100 100 100 100 100 100 100 100 Source: Grameen Bank Monthly Statements for October 1986 and December 1983, 1984, and 1985. One has to remember that the effective rate of interest is 3 0%. The GB charges 16% interest on its ®rounded off to the nearest whole number 178 t loans, the same as the commercial banks in Bangladesh. The GB member has to deposit 5% of the loan amount into the GF. Since the repayments are in equal weekly installments over one year, this means that the GB member saves 10% of the average amount outstanding for the year. In addition, 25% of the interest charges (that is, 25% of 16%) must be deposited in the EF. Thus, the effective rate of interest is 16% plus 10% for the GF plus 4% for the EF which equals 30%. It is noteworthy that savings takes precedence over increases in disposable income and consumption. Increases in the latter can occur only if the rate of return from the project exceeds 30%. It is convenient to consider the contribution into the GF as a form of forced saving and that into the EF as an insurance premium. 4.5.3 Impact of GB on its Members What has been the impact of GB on the lives of its members? Rahman and Hossain [1986] find that the average household income goes up from Tk 13,491 for first-time GB loanees to Tk 23,571 for fifth-time loanees. There is a strong and positive relationship between length of membership and income. Hossain [1986] finds that the amount of working capital per 179 loanee household increased nearly fourfold from Tk 743 to Tk 2,811 within a 2.25-year period between 1983 and 1985. Thus, not only did incomes rise (in spite of the high effective interest rates) but there is evidence of a spurt in productive investment. Given the low levels of income, this may appear surprising. But, as mentioned above, the savings threshold is likely to be fairly low in rural parts of LDCs. This, coupled with the tangible and immediate benefits newly attainable, makes saving and investment possible even at such low levels of income. This provides support for the hypothesis that poor people in LDCs are not irrational; given a choice, they save even out of their quite meager earnings to invest for a better future. The social and economic status of women has also definitely improved. Hossain [1986] finds that 55% of the loans disbursed and 69% of members are women. Rahman [1986] contrasts the situation of female loanees with women in the target group in a control village. The former buy 2 sarees (dresses) per year compared to the latters'1.65 and spend Tk 62 on medical expenses for housewives vis-a-vis the latters' Tk 17. Moreover- -and perhaps more significantly— 74% of the former 180 group voted in the last local election compared to the latters' 46% and 92% knew how to prepare oral saline versus 54% of the latter.9 These results demonstrate both an increasing awareness of, and participation in, the political process and enhanced knowledge about health care. Rahman [1987] finds that women are no longer ignored in family decision-making and wife-beating is virtually nonexistent among GB members. In a conservative, predominantly Muslim society, this newly found economic independence may be a harbinger of profound social changes of the type required for economic growth. There has been some opposition to the new role of women but these have been successfully overcome, as elaborated below. 4.5.4 Reasons for GB Success Why does the GB succeed? At the outset, a list of factors contributing to the success of the GB will be enumerated. Subsequently, it will be attempted to 9These data demonstrate, as noted in the next sentence of the text, that the women are participating more actively in national affairs and are more aware of health needs. While these changes may appear insignificant, it seems to me that they contain the seeds of social transformation that accompanies the process of economic development. 181 derive necessary and sufficient condition(s) based on the model developed in Chapter 3. Olson [1965] refers to the crucial role of the political entrepreneur in facilitating collective action. The unique personality of Yunus can hardly be overstated: his role is best conceived of as that of a benign, idealistic and self-sacrificing entrepreneur. It is doubtful whether the GB could have ever been started without the tireless efforts of such a person. The bank workers and officers— many of whom in the initial years were Yunus' students— are motivated and honest, especially in comparison to the staff of other banks. This dedication is universally documented as well as intriguing. Gibbons and Kasim [1990]— in a typical comment— state "... Grameen has become an island of honesty in the sea of corruption in Bangladesh." The job satisfaction that one gets from the rapid and visible positive effects on the poor people as a result of one's work appears to be very important. In terms of our model, an increase in one's psychic satisfaction, cH , may more than compensate for a decrease (or lack of increase) in one's monetary payoff, TMj. 182 At least six factors which favor collective action as identified by Nabli and Nugent [1989] are present in the GB scheme: (1) Small size of group: The primary group consists of five members who decide, by themselves, to form the group. This self-selection ex-ante makes it more likely that trustworthy people conducive to group success would be selected; ex-post. it is easier to monitor members' activities. (2) Homogeneous origin of members: Since all members come from the same village, they are very similar in terms of economic background and social norms and practices. (3) Length of association of members: In a typical rural society with very little mobility, members have known each other practically since birth and each others' families for generations. They are likely to have almost complete information about each other. This would help in two ways: i) in the initial phase of group formation, only those deemed likely to act on behalf of group interest would be selected and ii) the freerider problem is likely to be less acute since the members are very well known to each other. It needs to 183 be emphasized that these effects are independent of those in (l) and would be strengthened if the group were sufficiently small. (4) Close social and physical proximity of members: This factor would reinforce (3). If a member engages in conspicuous consumption— throwing a lavish wedding party for his/her daughter, for example— his/her action will not go undetected by the other members. (5) Sensitivity of the group to loss from inaction: The necessary condition for obtaining GB loan is the formation of the group. As discussed above, there are collective sanctions: if any member of the group defaults on a loan, the entire group becomes ineligible for the loan. (6) Cost of exit: As noted above, members' contribution to the GF are forfeited in the event of exit. According to GB annual reports of 1984 and 1985, the per capita savings in the GF and EF was Tk 1712 in 1985. Disaggregated data for the GF were not available but it is safe to speculate (given the formulae under which the two funds are set up) that it will amount to at least half this amount. This represents a sizeable amount of savings in the context of a poor rural 184 society. Hossain [1986] finds that 71.7% of the GB members had taken two or more loans. Thus, GB credit is not a one-shot affair but a continuous one. Several other factors contributing to the success of GB include: (7) The banking services are brought to the doorsteps of the people. Unlike the procedures for formal credit institutions, there are no costly and time-consuming travels to be made, no complicated forms to fill out and, perhaps most importantly, no requirement of collateral. The traditional literature on transaction costs often neglects sizeable noninterest costs of borrowing which include travel, entertainment, bribes, and the opportunity cost of the time involved in securing the loan. Fry [1988], quoting a survey, writes that the gross cost of borrowing from moneylenders averages 86% of the borrowed amount per year in rural Bangladesh whereas that of borrowing from commercial banks averages 108%. Recalling that the explicit nominal interest rate is 16%, the amount of noninterest cost is 92% per year. Since these 185 noninterest costs10 are virtually nonexistent in the case of GB, a major reason for its success is clear. (8) The rent-seeking behavior of bank officials needs emphasis: since formal credit is subsidized, bribe-taking by bank officials is common. Although hard to quantify, Hossain [1986] suggests that, in Bangladesh, a typical loanee has to pay to the bank manager "a significant proportion" of the loan amount as illegal "service charge". Casual empiricism suggests that the rate varies from 2% to 20%, with higher rates being applicable to smaller loans. Since, in the case of the GB, loans are approved jointly by the bank officials and members, such rent-seeking behavior is likely to be much less prevalent. Indeed, all surveys (for example, Hossain [1986], Rahman [1986], Rahman [1988], Gibbons and Kasim [1990] and Von Pischke [1991]) agree that GB officials are scrupulously honest. (9) Selective incentives in the form of the GF and the EF have been important for GB's success. As noted 10other than contributions to the EF and GF. Noninterest costs in this context refers to the transaction costs, including outright bribes. 186 earlier, one of the major problems of institutional credit is that it is not available for pressing consumption needs. Thus, there is a tendency for recipients to divert production loans to meet consumption needs or to resort to rural moneylenders who usually charge much higher explicit interest rates than institutional creditors. The result is often loan default and/or destitution. By means of the GF, the GB has created the same flexibility as the informal credit market. Members can borrow money for consumption at times of sickness or for social ceremony and do not have to either divert production loans or go to the informal sector. A form of social insurance has evolved: in 1985, according to Hossain [1986], 52% of the GF loans were used for nonproductive purposes like social and household needs, health and medical expenses, and repayment of outstanding loans. This attests to the fact that the demand for such loans exists. The EF can be used to repay the loan of a member who is unable to repay due to unforeseen circumstances like death or disability. This serves to reduce the riskiness of loans by creating risk-pooling. (10) Perhaps one of the two crucial reasons for the 187 success of the GB is the type of social norms prevalent in rural societies and the extent to which they are enforceable. In a rural setting, norms— such as, that it is essential to pay off all debts and to cooperate with fellow villagers— are strong and enforceable. Social ostracism is a credible and severe penalty in such traditional societies: no one, for example, may marry the daughter of a person who defaults on his/her loan or who acts in a manner detrimental to the group interest. In terms of our model, it needs emphasizing that such ostracism causes a reduction in private utility, Bj, because both the tangible payoffs, TM,-, and the psychic payoffs, cH , are reduced. The first occurs because of the sanctions imposed and the second because the noncooperator suffers psychic distress due to the divergence between his/her beliefs and his/her actions (or lack thereof). (11) In terms of by-products, the GB has started a major social program of sixteen goals to improve members* living standards. This program was designed in a way as to strengthen the norms conducive to GB ' success and that would increase mutual information. This is the second crucial element referred to above. 188 The goals emphasize self-reliance (that is, minimum dependence on government assistance), and the education of the members in sanitation, healthcare, nutrition, kitchen gardening of fruits and vegetables, family planning, and promotion of social reforms like dowry-free marriages. A measure of the seriousness with which these social programs are taken by the members is the finding by Hossain [1986] that 20% of the loanees could state all sixteen goals— a remarkable feat considering the fact that most of the loanees are illiterate; more than 8 0% could recite at least some of the goals. Hossain [1986] and Rahman and Hossain [1986] also find significant differences between GB members and eligible nonmembers in that the former grow much more fruits and vegetables, adopt contraceptive techniques more frequently, construct sanitary latrines more often, spend more for home construction, put more importance to educating their children, and, in general, display higher aspirations for themselves and for their children. In order to introduce unity and discipline, that is, increase control, C-, among members, physical 189 exercises and parades are held during weekly meetings. In a predominantly Muslim society where women rarely go out of their homes, one would expect such activities to be unpopular. The evidence is precisely the opposite: Hossain [1986] finds that 87% of loanees approve of these activities. Casual empiricism— newspaper reports and interviews, for example— suggests that there is a strong sense of pride among members, a sense of having achieved something without relying on government handouts. 4.5.5 How Does Our Model Help Explain GB Success What implications of the model developed in Chapter 3 are relevant here? It may be recalled that there is a twin problem in a Prisoners' Dilemma (PD) or PD-type situation (such as an Assurance Game): (i) one player does not have perfect information about the other's course of action and (ii) even if the players are allowed preplay communication, there is no assurance that any agreement arrived at will be honored. In such an uncertain, interdependent, decision-making process, we identified a probability threshold, q^- which is the minimum probability of j's 190 cooperation that makes it worthwhile for i to also cooperate, assuming risk-neutrality. The true qj, which is j's probability to cooperate, is obviously not observable to i. What does i do? (S)he forms an estimate, Pj, of qj, based on whatever information (s)he has about j. If Pj < q^, i will not cooperate. (S)he will cooperate with probability qf if p^ > qji. Why does i not cooperate with certainty when pj > q^-? This is because we are assuming self-interested individuals and if this was i's course of action, j would defect and leave i with the sucker's payoff. For similar reasons, j will cooperate with probability qj iff his/her estimated p, > qjj. We thus have: q , - = q, (Pj) q; = qj (P i) An equilibrium with expectations realized exists where: Pj = qj(P,) and P , - = q* (Pj) 191 The chances of cooperation are enhanced when the estimated pf is closer to the true qt - which means that expectations are realized. In case of the GB, the fact that the group is self-selected makes it more likely that this may be the case. The institutional setup of the GB helps to further increase the flow of information about each other. In this context, the weekly meetings are vital. Parades and physical exercises may appear trivial but they serve the absolutely vital role of collecting and disseminating information about each other. Thus existing bonds are solidified and new ones forged. In all traditional societies, there exists substantial control. The problem is to channel this control to gainful ends. If there is no private gain from some activity, then our model suggests that there is less likelihood of group cooperation. But there is no certainty of cooperation even if there are sizeable private benefits due to the PD or PD-type situation faced by the interdependent decision makers. The necessary condition for group cooperation (at least in a free society or for economic activities) is that there be private benefits. In terms of our model, it 192 means that Bf > 0. But this is not sufficient. A trivial sufficient condition is when there is total control, that is, when Ci = 1. Such a situation may be important in the study of political science or sociology but is not so in economic (or market) activity. What is required is the constant flow of information among group members. Institutions like the GB which facilitate this networking succeed in overcoming the basic reason for the existence of the PD. Once expectations are realized (in the first round of the game, for example), cooperation on subsequent rounds becomes easier. It is noteworthy that at these weekly meetings members not only discuss issues related to the GB but also their own experiences and problems, thus increasing the flow of mutual information. As an additional factor, the effect on internalized norms, c^., may be considered. We saw earlier that, ceteris paribus, anything that increased the psychic utility of an activity would increase the probability of cooperation. What is the probable effect of meeting the same group of people over an extended period of time at least once every week? It seems safe to speculate that internalized controls 193 would be strenghtened. It is also noteworthy that the weekly repayments are done publicly. This would increase the probability of repayment because, in the event of nonrepayment, the defaulter suffers psychic distress in the company of his/her peers.1 1 While a few other authors have attempted to analyze the reasons for GB success, none has focused on the solution to the information problem that the GB ultimately is. Varian [1990] models the GB situation as a principal-agent problem and tries to derive results for a situation in which agents monitor the performance of other agents. He does not make any explicit attempt to view the situation as one of strategic decision-making and the crucial role of information about each other in such situations. He does at one point say "... agents have a better information than the bank has about about who is a good credit risk." But he does not endeavor to model this lower information cost as significant in a strategic 1 1 As discussed in Chapter 6, emphasis on twice-a- month member meeting may well be a significant contributing factor for the success of the Women's Self-Employment Project in Chicago which is modelled along the lines of the GB. 194 situation and seems to view it merely as reduced transactions cost. There is, however, a more fundamental objection to Varian's model; it may not be correct to model the GB as a principal-agent problem at all. It will be recalled that loan applications are almost always approved on the basis of recommendation made by the center chief or the deputy center chief and that the GB participants are members. not mere clients as in other banks. Loanees have a substantial voice in how the GB works and, for all practical purposes, the GB staff only facilitate their actions. It is perhaps closer to reality to model the GB as an agent-agent (or principal-principal) problem rather than as a principal-agent problem. Braverman and Guasch [1987] cites three reasons for GB success; (a) the intertemporal linking of loans to induce repayment: no new loans are given until the old ones are repaid; (b) the use of strict auditing and accounting procedures and (c) joint responsibility of small groups of farmers with collective sanctions. The authors make no attempt to consider the role of norms in facilitating group formation. Nor do they discuss the importance of the dissemination of information in 195 strategic situations. Of the three reasons they cite, the first two are of a general nature and would contribute toward the success of any financial intermediary. The third recognizes the desirability of small groups and joint responsibility but does not seek to examine why these may be important. Authors like Varian and Braverman and Guasch make the crucial mistake of not attempting to incorporate the effects of norms on behavior. The model we have set out, on the other hand, does precisely that. It views group decision-making as interdependent (strategic) and recognizes the vital role played by information. It is not merely the small size of the group that is important as Braverman and Guasch [1987] seem to think but who forms the group (are they people known to each other?) and how the group is formed (is it self-selective?). It does not seem plausible that small groups would be equally likely to succeed, irrespective of whether they are formed in New York City or in rural Bangladesh. The neglect of the social context— history, culture, religion, norms— makes such analyses seriously wanting. 196 The importance of information has been emphasized, however, by Stiglitz [1990]. He argues that the GB provides an example of a successful peer monitoring system. The members of the peer group must be provided incentives to monitor others' actions. In the GB, this is done by denying credit to all members if any one defaults. Secondly, the small size of the group increases the risk from a single member's default and increases the incentives for peer monitoring. With large groups, there is the free rider problem: the costs to each resulting from a default are sufficiently small that the incentives to monitor would be minmal. Thirdly, there are strong incentives for the formation of groups with similar risk characteristics; a more risky member in a less risky group would in effect be subsidized at least ex-ante by the other members. Villagers have more information than formal credit institutions regarding the riskiness of potential individuals. Thus, the self-selection of the groups in GB is crucial in this regard. In terms of our model, 197 this helps ensure an equilibrium where expectations are realized.12 4.5.6 The Future of GB Will there be any problems as the GB continues to grow in size as it has been doing ever since its inception? If the primary group remains of small size (the present size of five members, for example), increase in the number of GB branches may not be a serious problem. The structure could be a federation representing the primary members as in many labor unions and chambers of commerce. This may even augment GB's bargaining power at the national level vis-a-vis other interest groups. Moreover, once an institution evolves and members receive tangible benefits, they are more likely to fight to keep it intact. In the case of the GB, initiating it was the most difficult part; given its huge success in the form of providing immediate and tangible benefits to its members, it should be possible to keep it going. Von Pischke [1991] thinks that the most important strength of the GB is in the area of choice of project 12See above. 198 and technology. The members select their own projects, thus determining both the size of the loan and the repayment obligation. Many credit programs in the LDCs promote a technology attractive to the donor or to the government. If this technology is not appropriate or successful, the borrower's commitment to repay the loan diminishes. The GB member, however, knows that even if her/his project fails, the group is committed to repay and the choice is made very carefully. It also seems logical that the borrower would have more information about the correct project than would a bureaucrat. One may expect there to be opposition to the GB from the commercial banks including the so-called Islamic banks. There is no evidence of such opposition and it is important to try to understand the reason(s) for this. First, the overwhelming popularity of the GB may render overt opposition to it more costly. One would then expect the other banks to take covert action like lobbying the Bangladesh Bank13 and/or the Finance ministry against GB expansion. Such activities have not been observed, either. The true reason for this 13The Bangladesh Bank is the central bank of Bangladesh. 199 lack of opposition is probably because the members of the GB would not qualify for loans in other banks nor would the banks be willing to go to the doorsteps of these rural people to solicit their deposits. Neither the commercial banks nor the "Islamic" banks have made concerted efforts to cater to the needs of the rural poor. Not surprisingly, there has been some opposition to the GB. The village moneylenders and political interest groups have lost respectively captive clients and henchmen. They have, however, found it difficult to openly oppose an overwhelmimgly popular institution. This is particularly the case because the GB is scrupulously neutral with respect to its political viewpoints. Opposition has taken the familiar and predictable form of using religious injunctions concerning the economic activities of women. Rahman [1986] recounts several such incidents of which two typical ones are discussed here. In one incident, the rural elite declared that women who joined the GB had become less religious and would not be allowed to cross the river by the ferry. The women solved the problem by organizing a separate 200 ferry. In another instance, a husband refused to allow his wife to continue to be a GB member in a village outside his own, even though his own village did not have a GB branch. He refused to live with his wife in the GB village and indulged in making disparaging remarks about GB and its role with respect to women. Forty women GB members converged on the man’s house and demanded that he seek a reasonable settlement of the issue with his wife. The upshot was that, within three days, the man apologized to his wife and created no further problems in the way of his wife’s continued GB membership. Obstacles of this type can be easily overcome if property rights are clearly defined and if there is a functioning judicial system. It should also be noted that problems of this nature are not unique to GB but affect any enterprise in the country. In Chapter 6, replicability of the GB— which appears more problematic— will be examined. 201 BIBLIOGRAPHY Adams, Dale W. 1978. "Mobilizing Household Savings through Rural Financial Markets", Economic Development and Cultural Change, vol. 26, no. 3, pp 547-60. Fry, Maxwell. 1988. Money, Interest and Banking in Economic Development. Johns Hopkins University Press, Baltimore. Furness, E. L. 1972. An Introduction to Financial Economics. Heinemann, London. Gibbons, David S. and Sukor Kasim. 1990. Banking on the Rural Poor in Peninsular Malaysia. Center for Policy Research, University Sains Malaysia, 11800 Penang Malayasia. Gonzalez-Vega, Claudio. 1983. "Arguments for Interest Rate Reform", in J. D. Von Pischke, Dale W. Adams and Gordon Donald (eds.), Rural Financial Markets in Developing Countries. The Johns Hopkins University Press, Baltimore and London. Hossain, Mahabub. 1986. "Credit for Alleviation of Rural Poverty: The Experience of Grameen Bank in Bangladesh", Bangladesh Institute of Development Studies, Dhaka and International Food Policy Research Institute, Washington, D.C. Lee, Warren F. 1983. "The Role of Financial Intermediation in the Activities of Rural Farms and Households", in J. D. Von Pischke, Dale W. Adams and Gordon Donald (eds.), Rural Financial Markets in Developing Countries. The Johns Hopkins University Press, Baltimore and London. 202 Nugent, Jeffrey B. and Mustapha Nabli. 1989. "Collective Action, Institutions and Development" in Nugent and Nabli (eds.), The New Institutional Economics and Development: Theory and Application to Tunisia. North Holland, Amsterdam and New York, pp 80-137. Ong, Marcia L., Dale W. Adams and I. J. Singh. 1976. "Voluntary Rural Savings Capacities in Taiwan, 1960-70", American Journal of Agricultural Economics, vol. 58, no. 3, pp 578- 81. Olson, Mancur. 1965. The Logic of Collective Action. Harvard University Press, Cambridge, Mass. Panikar, P. G. K. 1961. "Rural Savings in India", Economic Development and Cultural Change, vol. 10, no. 1. Rahman, Atiur. 1986. "Development, Awareness, and the Grameen Bank", National Institute of Public Administration, Dhaka University, December 4, 1986. -------- . 1987. "Dual Role of Tradition in Social Mobilization: The Grameen Bank and Rural Poor Women in Bangladesh", paper presented at the 21st Bengal Studies Conference held at the University of Wisconsin, Oshkosh, May 8-10, 1987. --------------- - 1988. "Loan Repayment Performance of Grameen Bank in Bangladesh", unpublished paper. --------------- and S. M. Hossain. 1986. "Impact of Grameen Bank Operations on the Level, Composition and Distribution of Income and Expenditure of the Rural Poor", Bangladesh Institute of Development Studies, Dhaka, Working Paper No. 5. 203 Rahman, Rushidan Islam. 1986. "Impact of Grameen Bank on the Situation of Poor Rural Women", Bangladesh Institute of Development Studies, Dhaka, Working Paper No. 1. Shaw, Edward S. 1973. Financial Deepening in Economic Development. Oxford University Press, New York. Stiglitz, Joseph E. 1990. "Peer Monitoring and Credit Markets", The World Bank Economic Review, Vol. 4, No. 3, pp 351-366. Varian, Hal R. 1990. "Monitoring Agents With Other Agents", Journal of Institutional and Theoretical Economics, 146: 153-174. Von Pischke, J. D. 1980. "The Pitfalls of Specialized Farm Credit Institutions in Low- income Countries", Development Digest, Vol. 18, No. 3, pp 79-91. ------------------. 1991. Finance at the Frontier. The World Bank, Washington, D.C. 204 CHAPTER 5 NORMS AND MOBILITY 5.1 Mobility and Control: Theory As discussed in Chapter 3, mobility is expected to have a negative effect on social control. In an immobile society, group members may know each other for years and are likely to have norms in common. Immobility also stimulates feelings of friendship and camaraderie which strengthen cooperation and facilitate the imposition of sanctions. Following Hechter [1984], any monitoring system has three aspects: (1) its metering capacity which depends on the degree to which the group has information with respect to compliance with expected behavior; (2) its sanctioning capacity which is a function of the resources it has for such activity; and (3) its capacity for allocation which depends on its ability to distribute these sanctions to reward compliance and/or punish non-compliance. In an immobile society, it seems plausible to hypothesize j that social control would be higher— due to improved monitoring and sanctioning— because of (1) and (3). 205 The information available to members of such groups would: (a) make it more likely that defectors can be detected and also (b) facilitate the allocation of reward/punishment. An additional point— overlooked by Hechter who does not model norms— but that needs emphasis is that any given sanction may have more impact. Consider the case of social ostracism which may be imposed as a punishment. This could take the form of group members agreeing not to have anyone marry a defaulter's daughter. Such a sanction could have negligible effects in a mobile, urban society but could have devastating effects in an immobile rural society. It will be recalled that, in Chapter 3, we used an Assurance Game (or PD-type game) rather than a pure Prisoners' Dilemma to capture this implicit willingness or propensity to cooperate. If, additionally, account is taken of the psychic suffering caused by alienation from one's peer group, it seems obvious why social control varies inversely with mobility. 206 5.2 Mobility and Control: Evidence Let us consider some examples of norms decaying (or weakening) over time to further support the hypothesis that increased mobility can be an important factor contributing to such decay/weakening. 5.2.1 Widows in India (1) The treatment of Hindu widows even a few decades ago can only be described as appalling. According to Basham [1989], for example, she had to live the life of an ascetic: sleeping on the floor, eating only one simple meal a day, wearing no perfumes or ornaments. Her sole duty was to devote herself to prayer and hope to reunite with her husband in the next life. Furthermore, she was inauspicious to everyone but her own children and thus was not allowed to participate in family festivals. Conditions were so bad that many scholars think that sati (self- immolation) was preferred over widowhood as a harsher but shorter ordeal. Altekar [1956] writes, "Many young widows found it a hard ordeal to lead a life of enforced celibacy and begun to prefer to die with their husbands, rather than to live (after their deaths)." 207 How was such a cruel norm maintained? It appears that two institutional setups which minimized mobility were crucial in this regard. The first was patrilocal residence which Dreze [1990] defines as the system in which "... a woman has to leave her parental home at the time of marriage to join her husband in his own village ... (and has) very little freedom to "return" to (her) parental home." The second is the restriction on widow remarriage which carries social stigma. A third factor is the patrilineal inheritance system. Dreze [199 0] finds that, in practice, widows have little inheritance rights. He distinguishes three land inheritance systems: (i) traditional law; (ii) modern law; and (iii) actual practice and correctly focuses on the third as the relevant one to study.1 Contrary to law, in practice if a widow has son(s), she does not 1 What this means is that the norm or the traditional system of land inheritance supersedes those the laws set out in the modern legal system. This is not uncommon in either developed or developing societies. The legal speed limit on most California freeways is 55 miles per hour but it is well known that the norm is that a highway patrolman will not give out speeding tickets until a speed of 65 miles per hour is i reached. Similarly, in many developing countries, ! dowry is banned and strict penalties are set out in the event of payment or acceptance of dowry; the norm, however, is that dowry is a common transaction. 208 inherit anything. The reason (excuse?) is that the widow may abandon her children and live with another man.2 If she leaves her husband's village (to remarry, for example) with her children, her children automatically forfeit their claim to their father's land. It is obvious that all three factors contribute to decreased mobility for the widow and thus make it easier for her to be treated in the ways described above. Dreze [1990], however, finds the condition of widows in rural West Bengal less oppressive than in other areas of India; land onwership, for example, is much more common than in Uttar Pradesh and Gujarat. It is interesting to try to understand why this is the case. The traditional land rights system in West Bengal is the dayabhaoa system while that in most of the rest of India is the mitakshara system. There is virtually no difference between the two except that, in the former, a widow could inherit even if she belonged to her husband's joint family at the time of the 2In a traditional kinship group or tribal society, ! this may have performed the role of endeavoring to keep the deceased's property from going to someone outside i the group/tribe. 209 latter's death.3 But this difference in the traditional law does not appear to be the crucial factor in the explanation of more extensive land rights of Bengali widows. The explanation for the difference in the condition of widows probably lies in the fact that in West Bengal there is movement away from traditional law toward modern law. Liberal reformers in the nineteenth century pioneered in legislative and social advances like the prohibition of sati. the abolition of child marriages, and the promotion of widow remarriage. These advances and a growing awareness among the female population have led to a movement toward modern land rights. Thus, we see that although traditional land rights persist in most parts of India, this is not so in West Bengal. The vital role is played by mobility. In West Bengal, the reformers have succeeded in bringing about changes that increased widow mobility while in other parts of India this has not occurred to the same degree. The position of widows in West Bengal is thus much better because | 3But, it needs to be pointed out that the i davabhaqa system protects the inheritance rights of ! widows somewhat better than does the mitakshara system. 210 the types of laws and customs restricting mobility and facilitating oppression of widows have been weakened. 5.2.2 Female Inheritance in Bangladesh (2) According to Islamic law, daughters get half the property of the father that sons get.4 But, in Bangladesh, where 86% of the population is Muslim,5 most women would waive their claim to the father’s property in return for appropriate dowry and an insurance motive. The latter would imply that, in the event that the woman should become a widow, the brother(s) would be expected to pay for the upkeep of her family until her children are grown up. If she had no children, she would usually stay with a brother until she remarried or died. It is noteworthy that the norm in this regard among Hindus is also very similar even though the Hindu religion does not have the clearly defined inheritance rights for women that Islam does. 4The Quran sets forth explicit guidelines regarding the inheritance of property in chapters ii, iv and v. 5This figure was obtained from the Worldmark Encyclopedia of the Nations. 1984. Vol. 4, Worldmark Press Ltd., New York. 211 Casual empiricism suggests that this norm is not as prevalent as in the past and that increased mobility that has accompanied economic growth is a major reason for the decay of this norm. In an immobile setting where people know each other intimately, a breach of the norm would lead to sanctions. Traditionally, a brother who refuses to care for his widowed sister would be condemned and ostracized. But, with increased mobility, the brother may live in the city and be immune to the sanctions, especially of the nonlegal variety like social boycott and ostracism. As a result, it is not surprising that more and more women are now asserting their "rightful claim"6 to the father's property.7 6that is, half the brother's share. 7This observation is based on personal knowledge about various people while living in Bangladesh as well as numerous discussions on this matter with friends and members of the family. An additional issue is: can a system of inheritance set down in the Quran (see footnote 3) for a tribal seventh century economy be valid without any modification to a modern twentieth century economy? In other words, some women question why they should get only half as much as a brother after the father's death. Note, however, that the inheritance right of women as set out in the Quran is extremely progressive for its times. We only need to recall that in the seventh century most European women had no property rights at all. 212 5.2.3 Jews in Brooklyn (3) Coleman [1988] cites the example of wholesale diamond markets where a merchant will hand over to another a bag of precious stones worth hundreds of thousands of dollars for the latter to examine, often without a formal receipt. "Th[is] ... market is Jewish, with a high degree of intermarriage, living in the same community in Brooklyn, and going to the same synagogues."8 If any member defects either by substituting or stealing stones, (s)he would lose family, religious, and community ties. The fact that the merchants live in the same geographic area makes it easier to detect defection and impose sanctions. With increasing mobility, elaborate bonding and insurance would be necessary; otherwise many mutually beneficial transactions would not take place. Here there is a similarity to the situation described by Nugent [1990]. The Jews who emigrated from various European countries to the U.S. tended to "settle in the same neighborhood" in big cities like l New York City. Consequently, they were able to hold on | 8emphasis added. Note that, in addition to lower I mobility, the relationships are multiplex. I 213 to their norms and customs from the Old World. In effect, it seems correct to say that they created "a home away from home". 5.2.4 Paternalism in the American South (4) Alston and Ferrie [1992] seek to demonstrate how a system of paternalism was instituted in the South to prevent the departure of blacks and how the system collapsed with the advent of mechanization. With the abolition of slavery, the most severe problem facing Southern agriculture was to assure an adequate supply of labor. Former slaves became free to leave an undesirable employer or even to move to other parts of the country. Furthermore, the hostility of poor whites toward freed blacks emerged in the open in the form of "whitecapping"— driving blacks from their homes and forcing them off the lands. The codification of such attitudes in the Jim Crow laws meant that "the entire machinery of the state became an instrument with which to coerce blacks." After emancipation wages rose either because (i) the labor supply curve had shifted to the left as workers sought more leisure or (ii) the labor demand curve had shifted to the right as more i I i workers were required to do the work previously done 214 under the onerous gang system or (iii) both. The result was an excess demand at below equilibrium wages rates. The planters turned to nonwage inducements like honesty, fair-dealing, and amenities like improved housing, garden plots, schools and churches. In addition, they took over the role of protector whereby they acted as intermediaries between their workers and the outside world. They posted bond for the workers, accompanied them to court, and stood surety for their workers' debts with local merchants. Thus, social control (in the form of racist laws) led to paternalism "... in which planters looked after most aspects of their workers' lives, and workers responded by offering their loyalty to their patron." A norm (paternalism) developed in response to social control sought to be imposed by law. It is important to note that this norm could only be nurtured if outmigration did not take place and that it evolved out of mutual self-interest. The black workers valued the protection from the pervasive lack of civil rights and society's condonement of violence that this form of paternalism offered; in economic terms, they preferred the nonwage 215 benefits such paternalism offered over the (additional) wage benefits they may have received. The above is an example of an interlinked deal in which "two or more interdependent exchanges are simultaneously- agreed upon."9 In many LDCs, a landlord takes on a tenant at a fixed rent and also agrees to provide him/her with credit at a fixed interest rate. It is observed in these countries that the rural interest rate (i) may be 100% while the urban interest rate (r) may be 10%. Does this mean that the rural folks are being exploited and why is there no arbitrage whereby the differential is reduced or eliminated? A proportion q of the loan is not repayed; it is convenient to think of a group of borrowers over whom the landlord-moneylender has control and can always recover his/her loans and another group from whom (s)he cannot. The expected earnings of the lender is: (1) (1 + i) (1 - q)L - L where L is the amount of the loan. The effective interest rate (d) is obtained by: 9Basu [1984]. 216 (2) d = i(l - q) - q Setting r = d, each landlord can charge the "equilibrium" i. The borrowers have an incentive to move to lenders who have no control over them. In such a situation, unless the lender discriminates between borrowers, (s)he will attract "lemons" and end up bankrupt. They would give loans only to those over whom they have control. The landlord-moneylender offers a package of wages and interest so that this control can be exercised. If the borrower-tenant wants to find a new lender, (s)he would need to find someone who would not only provide employment but also be a creditor.10 In the American South, a similar situation as the one described in the two paragraphs above is observed. The planters offered a package consisting of low wages and nonwage benefits which were important to the blacks. As in the case of the rural LDCs, control by the planters could be exercised and maintained only if ' 10In terms of our model, interlinkage would be an ' example of a multiplex relationship which facilitates i control (see Chapter 3 for details). 217 the deals were interlinked.11 Note that there is reciprocity in the relationship: the worker would not accept a low wage unless (s)he was convinced of the availability of the nonwage benefits.12 With the advent of mechanization, the excess demand for labor no longer existed. Because the production process was standardized, the cost of monitoring labor was also rendered easier. Black outmigration would no longer increase labor costs and threaten the Southern way of life. In the sixties, civil rights legislation was seen as a threat to white supremacy, a cornerstone of social control in the South. Thus, not only do we witness the demise of paternalism but legislative activity by Southern lawmakers to encourage black outmigration. Clearly, a norm (paternalism) which developed for mutual benefit I to counter the effects of social control (in the form I ' of laws and extralegal violence) tends to wither away 11The planter would not be willing to offer nonwage benefits (like standing surety for a worker's debt) in the event that (s)he could not pay low wages. 12For a survey of the interlinkage literature, see Bardhan [1980]. 218 when one side (planters) no longer perceives it to be beneficial. 5.3 Does Mobility Negatively Impact Control: Empirical Evidence 5.3.1 Background The main purpose of this chapter is to test an important hypothesis of the model developed in Chapter 3. It is that c^. or the control exercised by person j on person i depends negatively on the mobility of i and would depend on factors like relocation of residence and job change. The mean control C , - was defined as the average of the individual c^-'s, that is, the control exercised over i by all the other members of the group. What does C? mean in an operational sense? It is merely the amount of control i faces from the group on the average. Thus, if there are two people, 1 and 2, other than i in the group, C. would reflect the average control i faces from this group. In order to ensure that C. lies in the unit interval, it has been defined as: It may be the case that i cares about (or is controlled by) both 1 and 2. The reason may be psychic and/or tangible benefits. In other words, i's behavior is affected (but not determined) by 1 and 2. In particular, i is more likely to participate in a group which includes both 1 and 2 than in one that includes only 1 or 2, if and only if, i stands to benefit in some way, tangible or intangible, from such participation. The opposite may also be true; i may care about 1 but dislike 2. In such cases, i is more likely to participate in a group that consists of only 1 than in one that consists of both 1 and 2. Since the mean control, C; is the average of the ■ # individual controls or c^.'s, it depends functionally on the same variables that the c^. 's depend on. In the model set out in Chapter 3, Cjj and C{ were hypothesized to depend negatively on mobility, positively on the number of relations that group members have with each other, positively on the proportion of the group 220 cooperating, and in an indeterminate way on the number of people in the group. It is the purpose of this chapter to test empirically the hypothesized negative relationship between C? (or c^-) and mobility. This will enable us to understand not only why certain norms tend to persist but also to help devise policies that recognize the importance of the existing conditions. 5.3.2 cn and c^ : the difference and the relationship It needs to be emphasized that, in the computation of Ci, all Cjj's other than i = i. were considered. What does this mean? As discussed in Chapter 3, cn denotes internalized norms and beliefs. A person's private utility, B-, depends on both TMj (tangible or monetary payoff) and c^., thus: B, = Bf(TM,, c„) + + While TMf denotes a change in the payoff from the given utility function resulting, for example, from an increase in income, Cj. denotes a change in the utility function itself. This brings us to the crucial relationship between ci { and Cjj. It was shown earlier 221 that c^. tends to enhance cooperation but does not ensure it. Cooperation also depends positively on B}. It is thus certainly possible that in some cases cH would further strengthen the bonds of cooperation obtained from considerations of C5 (or c^.) and Bi along with Qi and Di which are, respectively, the group's mean threshold and defection probability without control. An example of the latter would be the case where an informal but effective old-age security scheme emerges. People not only find that such a scheme increases their satisfaction but they also believe that it is the right thing to do. What determines how c^- is formed is properly the territory of psychology or biology. It seems safe to speculate that some notion of fairplay or justice is almost certainly involved. Thus, in j societies without any formal social security scheme, it ! is perhaps not considered fair or correct to neglect i 1 one's parents who have sacrificed a lot and are now I j quite helpless. > I I It is also possible that c , - , and c^. could work in 1 opposite directions. An empirical example of this was [ seen in the Grameen Bank (GB) experience described in 222 Chapter 4. In many traditional societies— and especially in Muslim societies— the active participation of women in economic activities is frowned upon. What about a woman who does not believe that this restriction is fair? The C. would tend to influence the woman not to start her own business but the c^- would now work in the opposite direction. One of the major problems the GB faces is that some segments of rural society— the self-proclaimed mullahs, for example— are resisting what is essentially the decay of a norm. A similar situation was obsrved in many Western countries when women started to join the labor force. Many husbands thought that it was not "right" for the "weaker" sex to undergo the rigors of factory (or even office) work.13 This attitude was present even though it was recognized that, with the wife also working, the family would be better off financially. In the GB 13Reference may made, on a lighter note, to the television series called "The Honeymooners". In one episode, Ralph Kramden (played by Jackie Gleason) is furious that his wife Alice (played by Audrey Meadows) had decided to take a job to help supplement the family income. Ralph, in his own inimitable way, makes it very clear to Alice that it is the man's responsibility ! to earn for the family. 223 case, however, the opposition to women working does not come from their spouses but from outside "moral guardians". As our model would predict, with more and more women deviating from the norm, the norm has been disappearing very rapidly. Thus, in GB villages, women are not the subservient folk they are in many non-GB villages. What is important to note is not only the fragility of norms but also how beliefs and economic gains are intertwined sometimes in ways that are difficult (if not impossible) to decipher.14 It is almost certainly true that c^. responds to economic stimuli: the GB women were receiving tangible benefits. On the other hand, the process is perhaps not deterministic: it is not possible to say that the norm will decay if the costs of observing the norm exceed X dollars. And this is perhaps the greatest unknown: the influences of history, culture, religion. But a useful i i i -------------------------- | 14Nonetheless, it is imperative to endeavor to ! understand such phenomena. Easterlin [1978], for j example, criticizes a sociologist's observation that j values or attitudes are empirically unobservable and ; thus untestable residual. He refers to such attempts as "a semantic escape from one's ignorance" and 1 suggests that values or attitudes can be easily handled ; by varying the form of preference maps. 224 exercise in probabilistic terms is nonetheless possible. And this is what we seek to attempt here. We may not be able to say that a norm will definitely decay if X happens but if we are even able to predict that increased mobility (for example) is more likely, ceteris paribus, to cause the decay of a norm (like caring for elderly parent(s)), that itself may be of tremendous importance on both theoretical and empirical grounds. 5.3.3 The Data Set As noted above, our purpose here is limited to examining the effect of mobility on social control. Economists have rarely ventured into this field, leaving it to sociologists and anthropologists. Since there are tremendous data availability problems, it is important to ask how social control and mobility can be measured. Ideally, one would want to live for an extended period of time in the society one wishes to investigate and supplement quantitative data with qualitative information about belief systems, history, and culture. Time and monetary constraints did not allow this. Thus, we looked for a "second best" 225 solution: some data set which would allow us to quantitatively test at least some aspect of our model. The data set used for the testing of the hypothesis consists of eighty households and 559 variables. The respondents are a randomly selected set of permanent tea estate workers and a randomly selected control group. The data were collected by the Agro- Economic Research Centres in Assam and Santinitekan, India. The respondents were from the states of West Bengal (60) and Bihar (2 0) in India. The sample consisted of forty "tribal" and forty "non-tribal" people. Tribal people in the Indian subcontinent refers to the indigenous people usually living in the hilly or mountainous areas and having different language, culture and (in many cases) religion from the non-tribal people. An analogy would be the aboriginal population in Australia and New Zealand or the Native Americans in North America. The data show that 38 of the respondents were tea estate workers, 2 2 worked in agriculture, 9 were non- agricultural laborers and the remaining were craftsmen 226 or traders.15 40 of the 80 were Hindus, 20 were Muslims, 19 were Christians and one was a Buddhist. The proportion of Muslims and Christians is much higher than the national average (about 12% and 4% respectively).16 4 3 of the respondents were illiterate and 18 have had primary education with 2 being able to read and write only. The education level appears to be fairly typical for India where the illiteracy rate was 57% in 1985.17 48 of the respondents had incomes of Rupees 151 or higher per week.18 The per capita GNP for India in 1988 was US$34019 and thus the respondents are considerably better off than the national average. While 46 of the respondents have never moved since they started working, 3 0 did; in fact, 15 changed residence at least twice. Finally, the interviewees were a young group; 15For descriptive statistics, see Appendix B. 16Worldmark Encyclopedia of the Nations. 1984. Vol. 4, Worldmark Press, Ltd., New York 17World Development Report 1990, The World Bank, Oxford University Press, Oxford. 18US$1 = Rupees 15 19World Development Report 1990, The World Bank, Oxford University Press, Oxford. 227 all but 15 of them were in the age group 30 to 45. The oldest respondents were 50 years old. The reason for choosing this particular data set was that the questionnaire attempted to obtain information on not merely "economic" variables like income and consumption but also on matters like the role of parents in choosing their children's marriage partner, the number of times respondents changed residence, and whether respondents think their children will support them in old age. Ideally, we would have preferred time series data for the same households to be able to trace the process of norm decay/formation. But, given the purpose at hand, the data are adequate. In addition, as mentioned above, the questionnaire was administered to a cross-section of the population which, although not representative of the breakdown of the Indian population by occupation, meets our need of having a diverse sample living in a fairly remote area. It needs to be remembered that these people migrated from various parts of India and thus brought with them various cultural norms which, over the course of time, interacted with each other in the formation of new norms. 228 The analysis that follows is a partial-equilibrium one in that we are holding constant Cl (closure), w (the proportion of the group cooperating), and n (the number of members in the group). The reasons for doing so are: (a) non-availability of data and (b) since we are studying one group, the inter-group differences attributable to Cl, w and n can reasonably be assumed to be constant at one point in time. 5.3.4 Problems of Data Availability To measure Cl, for example, one would require information about the type and extent of relationships between between members of the group. As discussed in Chapter 3, closure exists when, for example, two persons B and C have relations with each other such that they can combine to sanction another person A. Thus, when A considers breaking an obligation toward B (non-repayment of loan, for example), the existence of closure may significantly affect A's propensity for such (non)action. Participation in multiplex relationships— fellow worker, coreligionist, parents whose children go to the same school— would tend to increase the extent of control. One could measure closure by first identifying the total number of 229 possible (or important) interpersonal relationships within the group and then considering the ones existing. The ratio of the existing to the possible relationships would provide an index of the extent of closure in the group. This would require data which have sociological, cultural, and anthropological information and such data were not available. To measure w, one would need subjective questions about the respondents' views on various norms and objective cross-checks on such views. This needs to be done because, as pointed out by Kuran [1987], people's publicly expressed opinions often vary significantly from their private preferences and courses of action. Thus, a respondent may state that (s)he finds the caste abhorent but his/her actions may not coincide with the expressed opinion. As discussed in Chapter 3, the crucial role in the dynamics of norm formation and decay is perhaps played by w, the proportion of a group cooperating. It is a predetermined variable because it enters a certain period with a certain value. But, whenever the net number of cooperators (that is, new cooperators minus new defectors) is different from zero, w will assume a new value at the end of the time 230 period. Not many years ago, the idea of not having women serve in the armed forces would have had overwhelming support. Considering the entire country as one group, w would have been high. Over time, however, w has fallen and this norm has decayed. To measure change of this type, time-series data— where the same respondents are questioned year after year— are required. Data availability problems restrict us to a cross section analysis which is equivalent to a freeze frame in a motion picture. Finally, with respect to n, since we are studying only one group consisting of eighty households, n in this case is eighty and constant. We are thus using this sample as an approximation for a typical rural society in India. Like most rural Indian societies, a tea estate is a largely self-contained community. Even though there are significant differences among villages in rural India, the similarity of the norms across ethnic and religious lines is remarkable. An example is the norm regarding caring for elderly parents which is prevalent among Hindus, Muslims and Christians and among rich and poor alike. If we can show that increased mobility is more likely to lead to reduced 231 control, it is reasonable to conclude that the same is likely to be true for rural society in general. If we wanted to investigate the differential impact of n, we could find study groups of different sizes and then observe the resulting differences in control as n varied. 5.3.5 How to Measure Control After selecting the data set, the next task is to find a way of measuring some form of social control. As noted above, it would be ideal (but was not feasible) to supplement data obtained from questionnaires with an on-the-field evaluation of beliefs and norms. For the task at hand, we searched for some variable(s) that would serve as a proxy for (some form of) control. In other words, the variable(s) would be such that one agent's behavior is affected by other(s). Three candidates suggested themselves: (1) whether respondents had influence over the choice of children's marriage partners; (2) whether respondents think that children will be supportive of them in their old age; and (3) whether respondent or other siblings provide old age support for their 232 parents. Each of these variables could serve as a reasonable way of measuring control.20 The final choice among these was essentially due to the nature of the data. With respect to the first candidate variable, there was in fact little variation in the variable. Specifically, the respondent was asked about the method of selection of marriage partner for (i) male and (ii) female offspring. There were six possible answers, namely, "arrangement by": (a) "respondent" (b) "respondent's spouse" (c) "the couple" (d) "other person" (e) "children" and (f) "other". If we interpret (a), (b) and (c) above as responses demonstrating parental control over their children's marriage partners, the data show that, in 70 cases, parents had such influence over their sons' choice of wives-to-be. There were only 7 cases of the sons selecting their own life-partners and there were three missing values. For daughters, the situation is 20Npte that we call this control Cf and not cn which is internalized norms or beliefs. To study the latter, we would require subjective answers as to what the respondents think. I have used control and norm synonymously in this chapter and it is important to remember this crucial distinction between a social norm and an internalized one. 233 similar: there are five missing values, seven cases of self-selection of husbands-to-be and the remaining 68 cases were instances of parental choice. It is interesting to note that the seven cases of self- selection of marriage partners by both male and female children occurred in the same households. This is perhaps an example of a situation of the norm "let parents choose marriage partners" having broken down. The lack of variation renders parameter estimation meaningless. To measure the second candidate variable properly would require information from the children, perhaps to be augmented by time series data. While there was some information about the children, the primary focus was on the household head and spouse. No time series data were available. Our choice of the third variable as a empirical proxy for social control was therefore largely by default of the other two. This, however, does not mean that there are not strong theoretical grounds for the choice. In many traditional societies, caring for old parents is a strongly ingrained norm. In the absence of old-age pension or social security, such behavior is 234 often the only support available to elderly people. This norm of children caring for their parents in old age is passed on from generation to generation and is an example of what Ellickson [1991] calls informal control "that arises out of the operation of decentralized social forces". Casual empiricism suggests that, in all pre-modern societies, this is an almost universally prevalent and powerful norm. Given that the data allow us to quantify— albeit in binary form— the existence of this norm, the logic behind using it as a proxy for social control is clear. What we are doing is examining how an existing norm— children should care for their old parents— is affected by variables such as mobility. Caring for elderly parents captures effective control over long periods of time. As noted above, the tradition is handed down through generations. In countries like India, loyalty of children is common (especially in villages) though by no means universal. This is consistent with the model developed in Chapter 3. The presence of a norm merely makes a particular behavior more likely; it does not render it a certainty. As the private costs of caring for one's 235 parents rises, one is likely to observe the withering away of the norm. In urban areas where living space is more limited and where caring may not be shared by other siblings, these costs are likely to be higher and it is not surprising that we observe this phenomenon less frequently. In addition, one has to consider the extra monetary costs of moving one's parent(s) and the diminished likelihood of social sanction (e.g., gossip and/or ostracism) in an urban setting. The first increases monetary or tangible costs of bringing one's parent(s) to live with one's family. The second renders the intangible or psychic costs of not doing so less costly. As shown in Chapter 3, this implies a reduced likelihood of undertaking the action contemplated. The process of urbanization is almost everywhere synonymous with the process of economic development and mobility with urbanization. If increased mobility implies even a partial breakdown of the norm of caring for elderly parents, the policy implications could be profound. This is especially true if the rate of change of mobility is expected to be much faster than the historical trend (which is 236 quite possible given the tremendous advances in transportation and communication and the profound political changes of the past several years). 5.3.6 How to Measure Mobility The available data set for rural India included three possible measures of mobility: (1) number of years in present place of employment; (2) number of years in present house; and (3) number of times respondent changed residence. Measure (1) is not in general a good measure of mobility, especially in a rural society. A farmer could be working for various people during a year, yet if (s)he were rather rigidly tied to one of these employers at one particular in time, (s)he would not necessarily be mobile. Thus, a person could be employed only for a short time at the place of employment when the survey is conducted but that would not necessarily be a good measure of his mobility. Both measures (2) and (3) appear on theoretical grounds to be reasonable measures of mobility. Preliminary investigation with various forms of model specification involving both (2) and (3) (separately) as independent variables, however, showed that with measure (2) as the independent variable, the 237 results were extremely sensitive to changes in model specification. Thus, we chose (3) as the empirical measure of mobility. 5.4 The Estimation Procedure and Results Probit was used to estimate the parameters.21 Given that the dependent variable was a zero-one dummy variable— 0 (when neither the respondent nor any other sibling was caring for an old parent) and 1 (when some sibling was caring for an old parent), a probit estimation procedure was used. Specifically, it is convenient to consider the numerical value of the dependent variable as an indicator of the probability of the event occurring. The variables X68 and X69 measure whether the respondent or any sibling takes care of an elderly father or mother respectively. If they do, the variable equals one, and, if they do not, it equals zero. The estimated X68 and X69 can thus be 21See Appendix A for a brief description of the probit technique. 238 considered as the probability of the event (caring for elderly father by children, for example) occurring.22 5.4.1 Carina for the Elderly Father The estimation results reported below support the hypothesis that more mobility implies less control. The variables are defined below: X68 dummy variable measuring whether the respondent or any siblings takes care of, or ever took care of, the father N83 the number of times the respondent changed residence divided by the respondent's age VB64 whether the father is alive VB65 age of the father SQVB65 square of VB65 Descriptive statistics for the above variables as well as for those relating to the mother are provided in Appendix B. As expected in a rural setting, most elderly fathers (about 80%) and elderly mothers (82%) are looked after by an offspring. Note the low mean 22This interpretation is appropriate in the present case for living parents only. 239 value of N83 which is not surprising; perhaps more significant is the high standard deviation. Thus, although the community is fairly immobile, there is substantial variation around the mean. The probit results obtained (with X68 as the dependent variable) are; Model 1 Parameter Estimate Error T-Statistic C23 1.0432 .205 5.0885 N83 -11.928 5.311 -2.2458 R-Square = .107870 The result is consistent with our hypothesis that mobility would have a negative effect on control. Indeed, the negative coefficient is significantly different from zero at the 5% level. It was then decided to experiment with various specifications both to test the sensitivity of the parameters to changes in model specification and perhaps to increase the explanatory power of the model. In all specifications, our measure of mobility, N83, 23C denotes the constant term. 240 was included and thus it was possible to examine the effect of other explanatory variables while holding N8 3 constant and, furthermore, to investigate whether the effect of N83 was sensitive to changes in specification. Thus, we added living status of the father, VB64. This variable would be expected to have a positive effect on the giving of support. As shown by the following result, this expectation is confirmed, but the coefficient of VB64 is significant only at about the 10% level. It is also noteworthy that the explanatory power of the model is increased only marginally by the inclusion of VB64. Model 2 Parameter Estimate Error T-Statistic C .8703 .22667 3.8396 N83 -11.804 5.0131 -2.3546 VB64 .65812 .41803 1.5743 R-Square = .128327 Next we investigate the effect of the age of the father. In the original dataset, VB65 was given a missing value if the father was dead. There were 55 missing values and 25 living fathers. If we deleted 241 all missing values, we would lose a lot of information. What we decided to do was to compute the mean of the living fathers and use that as an approximation for the age of the dead fathers. The mean of the 25 living fathers was about 62 years and we decided to use 65 years as the mean of the dead fathers (assuming that they lived a little longer than the surviving fathers). VB65 would be expected to have a positive effect on X68. Our finding below does not, however, confirm this. Model 3 Parameter Estimate Error T-Statistic C 4.12122 2.43021 1.69583 N83 -12.7225 5.72587 -2.22194 VB65 -.047521 .037195 -1.27761 R-Square = .119903 It is noteworthy that VB64 is more significant than VB65, perhaps suggesting that age adds nothing to living status as the explanation of the decision of whether or not to care for one's father. Note also that the R2 of Model 3 is less than that of Model 2, 242 that is, the explanatory power of the model is not increased by adding VB65 and dropping VB64. With both VB64 and VB65 included, one would expect both variables to individually affect X68 in a positive way. The results obtained do not confirm this hypothesis: Model 4 Parameter Estimate Error T-Statistic C 3.50432 2.61343 1.34089 N83 -12.3447 5.37423 -2.29702 VB64 .597171 .431633 1.38352 VB65 -.040337 .039785 -1.01388 R-Square = .131334 Note that the sign of VB65 has remained the same. This may be attributable to the collinearity between VB64 and VB65. In the original dataset, if the father was dead, VB65 was recorded as a missing value. But, as noted above, VB65 was recoded to equal 65 years when the father was dead. Thus, for many observations whenever VB64 was equal to zero, VB65 was equal to 65. 243 As shown in Appendix C24, however, the correlation coefficient between the two variables is -0.27856 which does not appear so high as to cause estimation problems. The second reason may be due to the presence of non-linearities. It was decided to drop VB64 and introduce SQVB65. It is expected that VB65 will have a positive effect on X68, that is, 6X68/5VB65 > 0 in the relevant range of values of VB65. The result presented below confirms this expectation: Model 5: Parameter Estimate Error T-Statistic C -5.46585 12.955 -.421911 N83 -12.8249 5.72663 -2.23951 VB65 .247309 .393644 .628255 SBVB65 —.225403E-02 .299896E-02 -.751603 R-Square = .141230 The signs of the parameters are of the expected kind. Note that 6X68/6VB65 > 0 for all VB65 > 54.8 years, that is, the likelihood of providing old age 24Table A 244 care increases with the father's age. It needs, however, to be noted that the age variable is not statistically significant. Further attempts were made to improve on the model by adding other variables like income and years of participation in provident fund; neither of these were significant. The reasons are perhaps: (i) the probability of caring for one's father is not affected by one's income: it may have been interesting if the father's income figures were available; (ii) respondent's participation in provident fund is probably not particularly important in affecting his decision to care for his father: once again, the crucial variable may be the father* s access to provident funds the presence of which may lower the respondent's proclivity to care for his father. 5.4.2 Carina for the Elderly Mother Similar investigation was carried out to test the likelihood that the respondent or any of his siblings would care for his old mother. The new variables— analogous to those for the father— are: 245 X69 dummy variable measuring whether the respondent or any sibling takes care of, or ever took care of, the mother VB56 whether the mother is alive VB57 age of the mother (with a similar transformation made as in the case of VB65, that is, if the mother is dead, VB57=68)25 SQVB57 square of VB57 The results obtained when the dependent variable is X69 are strikingly similar to those for the father, X68. There are, however, some differences as discussed below. 25It will be recalled that the mean age of surviving fathers was about 62 years. The mean age of surviving mothers was about 59 years. It was decided to assign an age of 65 years (62 + 3) for all deceased fathers. It did not appear correct to use the same algorithm and assign an age of 62 years (59 + 3) for deceased mothers. This is due to at least two reasons. The age difference between husband and wife is high in India, often more than ten years. While the mean age of surviving mothers is less than that of surviving fathers, this does not necessarily mean that the former die at an earlier age. According to the 1990 Demographic Yearbook, the life expectancy at 3 5 years was 32.6 years and 35.4 years for men and women respectively. Thus, it was decided to impute an age three years higher for the mothers (that is, 65 + 3 = 68 years). 246 Model 1 Parameter Estimate Error T-Statistic C .99204 .19766 5.0188 N83 -5.4938 4.4974 -1.2216 R-Square = .311224E-01 It is noteworthy that, although N83 is still negative, it is significant only at the 20% level compared to the 5% level in the case of the father. Furthermore, the estimated value of the coefficient of N83 is smaller (in absolute terms) than that of N8 3 with X68 as the dependent variable. The negative impact of mobility on the control variable or norm is thus less likely and, assuming it occurs, less strong for the mother than for the father. This may suggest that the norm for old age care is stronger for the mother, reflecting perhaps the fact that the mother in most rural societies is deemed more helpless than the father. As in the case of the father, we held N83 constant and added other expalnatory variables both to test the sensitivity of the parameters to changes in model specification and to attempt to increase the explanatory power of the model. We added the living 247 status of the mother, VB56. The result below shows that, as expected, it has a positive effect on the provision of support but that it is significant at about the 10% level. Model 2 Parameter Estimate Error T-Statistic C .80207 .22544 3.5579 N83 -7.5917 4.7591 -1.5952 VB56 .6304 .38446 1.6397 R-Square = .601000E-01 Next, we tested the effect of the age of the mother. Analogous to the case of the father, VB57 has been redefined to equals 68 if the mother is dead and the mother's age if she was alive.26 The result shows that VB57 is less significant than VB5627, perhaps suggesting that age is not an important consideration in decisions regarding care for the mother. Note that N83 is also less significant in the specification which includes VB57 compared to that which includes VB56. 26see above footnote for details 27In fact, VB57 is not statistically significant. 248 Model 3: Parameter C N83 VB57 Estimate Error T-Statistic .451523 1.52326 .296418 -5.18225 4.5635 -1.13559 .833563E—02 .023323 .357401 R-Square = .341846E-01 With both VB56 and VB57 included, the result is: Model 4 Parameter Estimate Error T-Statistic C -6.42891 4.00262 -1.60617 N83 -7.60238 4.68896 -1.62134 VB56 2.00193 VB57 .10634 1.00258 1.99678 .05886 1.80665 R-Square = .896937E-01 Note that VB56 is now more significant than it was in Model 2. VB57 goes through an even more radical transformation: from the insignificance in Model 3 to the significance (at almost the 5%) level in Model 4. What do these results mean? When a slight change in specification leads to big changes in parameter estimates, it is necessary to be careful in 249 interpreting the estimation results. Note that, in Model 2, the R2 is higher than that in Model 3 and also that N8 3 is more significant in Model 2 compared to Model 3. Furthermore, as noted above, VB57 is not significant in Model 3 while VB56 is significant at about the 10% level. It seems safe to hypothesize that VB56 is the more important explanatory variable while VB57 is not. It may seem that the most likely reason for this dramatic change in the estimate of VB57 from Model 3 to Model 4 is due to the collinearity between VB56 and VB57. As noted above, this is because VB57 equals 68 for all cases where VB56 equals zero. In fact, the correlation coefficient was high (-0.59424)28 but does not appear high enough to cause severe estimation problems. The other theoretical possibility is the presence of nonlinearities in VB57 which we next decided to investigate. It was then decided to drop VB56 and introduce SQVB57. It is expected a priori that that the effect of VB57 on X69 will be positive. The result obtained 28Appendix C, Table B 250 shows that VB57 is not significant and that its positive impact on X69 is at an age which may be considered outside the relevant range.29 Model 5: Parameter Estimate Error T-Statistic C -2.99752 9.82174 -.305192 N83 -4.98052 4.54803 -1.09509 VB57 .125593 .330338 .380195 SQVB57 .977851E—03 .274467E-02 -.356272 R-Square = .344776E-01 6(X69)/8(VB57) > 0 for VB57 > 64.02 years. This may well be considered outside the relevant range and is much higher than that obtained for the father. It is important to note that neither VB57 nor SQVB57 is remotely statistically significant. Furthermore, comparing the R2 in Model 1 and with that in Model 5, it is obvious that age does not add significantly to the explanatory power of the model. It seems reasonable to conclude, therefore, that age is not an 29see below (next sentence) 251 important consideration in determination of X69. As in the case of the father, the addition of other explanatory variables like income or years of participation in provident fund had no significant impact on X69. 5.4.3 Some Concluding Remarks In general, the results confirm the hypothesis that mobility reduces control as measured by the likelihood of caring for elderly parents. They also demonstrate that mobility has smaller and less certain effects on caring for the mother than it does for the father. Caring for the father may involve considerations of (expected) tangible gains in the form of (an increased probability) of a favorable bequest. Caring for the mother, on the other hand, imparts psychic or intangible gains.30 What the results suggest is perhaps that, when children view other tangible benefits which increased mobility can obtain for them (a job in the city, for example), their strategic 30This surmise is made in view of the fact that property in rural India is almost exclusively owned by the male household head. 252 % behavior vis-a-vis their father with respect to bequest is affected. In particular, they may view the (tangible and perhaps more certain) gains of such employment preferable to the (expected) tangible gains from a (future) favorable bequest. The above details are interesting but are not crucial for our purposes. What is noteworthy is that no matter which dependent variable— X68 or X69— we consider, or which model specification we examine, mobility always has a negative impact on control and that it is almost always statistically significant at conventional levels (5%). This provides strong empirical support for the hypothesis of our model that social control is likely to be weaker in more mobile societies, ceteris paribus. The sixth and final chapter will deal with the implications of the findings in Chapters 4 and 5 and will suggest new fields of research in the light of these findings. It will be shown how mimicking of the Grameen Bank in Malayasia proved a dismal failure at the outset because of different objective conditions. The importance of informal old-age security in societies where formal retirement schemes are almost 253 non-existent is obvious. Given the priorities and budgetary constraints faced by most developing countries, it is unlikely that such formal schemes will be launched on a wide scale any time soon. The dimensions of the problems of old-age care can be appreciated if one examines the available data. Tracy [1991] cites estimates31 that, of the projected world population of 1.2 billion in 2025, 71% will live in the developing world and the age dependency ratio32 will rise from 7% in 1950 to 12% in 2025. According to another United Nations source,33 as fertility declines and as life expectancy increases, the highest growth of population will occur among the very oldest, that is, those 80 years and older. While the populations 60 years and older and 70 years and older are projected to grow by 57% and 63% respectively over the period 1985- 31from Economic and Social Council. 1989. Second review and appraisal of the implementation of the international plan of action on aging. Report of the Secretary-General (E/1989/13). United Nations, New York. 32that is, (population aged 65 or more divided by population 15 to 64) X 100 33United Nations Fund for Population Activities. 1985. Population Perspectives: Statements bv World Leaders. Second Edition. New York. 254 2000, the population 80 years and older is expected to grow by 82%. Furthermore, Treas and Logue [1986] estimate that life expectancy differences at age 60 between developed and developing regions is only 2.4 years for men and 4.6 years for women. This means that, for persons who survive to age 60, societies and/or families may have to provide old-age support for about the same number of years regardless of the level of social and economic development. Tracy [1991] provides four reasons why social security schemes like those in the developed countries (DCs) may not be viable in the less developed countries (LDCs). First, most LDCs lack the social and economic prerequisites for such schemes. Governments lack the capacity to collect taxes and dispense benefits and the proportion of population engaged in full-time wage employment is not sufficient to support programs of social insurance. Secondly, there is the perception that if the DCs face continuing problems in financing their social security schemes, the LDCs with significantly fewer resouces are unlikely to be able to replicate DC-type schemes. Thirdly, there is the question of cultural divergence: many LDC scholars 255 think that the social security schemes in the DCs have resulted in conditions unacceptable to the cultures of the LDCs. Finally, many LDC scholars believe that the DC goal of redistributing responsibility of elderly care to families and communities while minimizing government intervention is something that the LDC already has and is in danger of losing if DC-type schemes are introduced. Tracy [1991] concludes that the dilemma is "... how to offer support through federal and state programs without destroying the traditional responsibility of families to care for their elderly members." Thus, it is vital that the informal schemes are nurtured with the correct economic incentives and with an effective legal system at least as a (perhaps fairly long) temporary alternative. This is precisely the point made by Nugent [199 0] when he says "... the socially efficient solution to the problem of that interim period is to foster the forces of cooperation within extended families and to extend the longevity of the traditional system." He advocates two means of accomplishing this goal: (1) the use of bequests as a strategic instrument to sanction noncooperation such as children not taking care of 256 their elderly parents and (2) maintenance of a reciprocal norm whereby parents invest in children and the latter repay by caring for the former in their old age. In terms of our model, (1) would increase the private (expected) utility from cooperating, Bf, and would thus facilitate cooperation. In view of the results obtained above, (2) is more problemmatic. If there are situations, as Nugent [199 0] describes, where villagers in Mexico and Peru move to the city and settle in the same neighborhood, the flow of information among them is not seriously impaired and thus cooperation may not be greatly affected. But it is not clear that this is the typical situation in other developing countries where norms may wither away.34 Our model would then suggest that private benefit35 be increased for cooperative behavior in addition to taking measures to increase the flow of information. 34Nugent [1990] considers mobility given and seeks ways of mitigating its influence on the norm. 35This could take the form of tax breaks and/or subsidies. An example would be allowing deduction for an elderly parent from one’s income tax or exemption from agricultural land tax or access to subsidized food (via the rationing system, for example). 257 BIBLIOGRAPHY 1990 Demographic Yearbook. 1992. United Nations, New York. Alston, Lee J. and Joseph P. Ferrie. 1992. "Paternalism in Agricultutal Contracts in the U.S. South: Implications for the Growth of the Welfare State, unpublished paper. Altekar, A. S. 1956. The Position of Women in Hindu Civilization. Motilal Banarsidass, Delhi, India. Bardhan, P. K. 1980. "Interlocking Factor Markets and Agrarian Development: A Review of Issues", Oxford University Papers, 32. Basham, A. L. 1989. The Wonder that was India. Rupa Press, Calcutta, India. Basu, Kaushik. 1984. The Less Developed Economy. Basil Blackwell, New York. Coleman, James S. 1988. "Social Capital in the Creation of Human Capital", American Journal of Sociology 94 Supplement: S95-120. Dreze, Jean. 1990. Widows in India. Development Economics Research Programme, Suntory-Toyota International Centre for Economics and Related Disciplines, London School of Economics, DEP No. 26, August 1990. Easterlin, Richard A. 1978. "New Directions for the Economics of Fertility", in J. Milton Yinger and Stephen J. Cutler (eds.), Major Social Issues: A Multidisciplinary View. The Free Press, New York. Ellickson, Robert C. 1991. Order without Law. Harvard University Press. Cambridge, Mass. Gujarati, Damodar N. 1988. Basic Econometrics. McGraw-Hill, New York. 258 Hechter, Michael. 1984. "When Actors Comply: Monitoring Costs and the Production of Social Order", Acta Sociologica 3: 161-83. Kuran, Timur. 1987. "Preference Falsification, Policy Continuity and Collective Conservatism", Economic Journal 97: 642-665. Maddala, G. S. 1989. Limited-Dependent and Qualitative Variables in Econometrics. Cambridge University Press, Cambridge. Nugent, Jeffrey B. 1990. "Old Age Security and the Defense of Social Norms", Journal of Cross- Cultural Gerontology, Vol. 5, No. 3. Tracy, Martin B. 1991. Social Policies for the Elderly in the Third World. Greenwood Press, New York. Treas, Judith and Barbara Logue. 1986. "Economic Development and the Older Population", Population and Development Review, Volume 12, Number 4, pp. 645-673. 259 APPENDIX A The probit technique is summarized in Maddala [1989]. It is assumed that there is an underlying response variable y* defined by the regression relationship: (1) y.* = b ' Xj + u, where y* is not observable. What is observed is a dummy variable y such that: (2) y = 1 if y,* > 0 and y = 0 otherwise Thus, E(y,*jxi) = b 1 Xj. From (1) and (2), we obtain: (3) Prob(yj = 1) = Prob(ui > -b'x,-) = 1 - F(-b,xi) where F is the cumulative distribution function for u. The observed values of y are vary from trial to trial (depending on x}) and are realizations of a binomial process with probabilities given by (3). The probit model assumes that the cumulative distribution of u , - is normal. Another model— the logit— assumes that the cumulative distribution of uf is logistic. The normal and logistic distributions are very close to each other, except at the tails. Thus, unless the sample size is very large so that there are 260 many observations at the tails, the two techniques would be equivalent. As Gujarati [1988] writes, "[t]he choice between the two is one of (mathematical) convenience and the ready availability of computer programs." In our example, since the sample size was eighty, the choice between the logit and probit models was determined by the availability of probit estimation on TSP. 261 APPENDIX B DESCRIPTIVE STATISTICS MEAN STD. DEV. MINIMUM MAXIMUM X68 .79221 .40839 0 1.0 VA10 39.73750 6.55018 30.0 50. 0 VB83 .76316 1.31523 0 8.0 N83 .019728 .035062 0 .22222 VB64 .3125 .46644 0 1.0 VB65 64.1375 4.52768 50 81.0 SQVB65 4133.8625 586.75183 2500 6561.0 X69 .81818 .38822 0 1.0 VB56 .4125 .49539 0 1.0 VB57 64.3625 7.10721 45 80. 0 SQVB57 4192.4125 859.80616 2025 6400.0 VA7 1.76250 .86043 1.0 4.0 VA11 2.16250 1.36357 1.0 5.0 VA12 1.0 0 1.0 1.0 VA13 2.67500 1.94074 1.0 7.0 VA175 5.65000 1.77963 1.0 8 . 0 262 X68 dummy variable measuring whether the respondent or any sibling takes care of, or ever took care of, the father VA10 age of respondent VB83 number of times respondent changed residence N83 VB83 divided by VA10 VB64 whether the father is alive VB65 age of the father SQVB65 square of VB65 X69 dummy variable measuring whether the respondent or any sibling takes care of, or ever took care of, the mother VB56 whether the mother is alive VB57 age of the mother SQVB57 square of VB57 VA7 religion: 1 for Hindu, 2 for Muslim, 3 for Christian, 4 for Buddhist VA11 education: 1 for illiterate, 2 for 3Rs, 3 for primary, 4 for middle, 5 for matriculation, 6 for higher secondary, 7 for college graduate, 8 for masters and 9 for higher than masters VA12 marital status 263 VA13 economic status: 1 for permanent tea estate worker, 2 for casual tea estate worker, 3 for agriculture worker, 4 for craftsman or artisan, 5 for non-agricultural worker, 6 for businessman or trader, 7 for service, and 8 for student or non earner VA175 average weekly income (in rupees): 1 for 0-50, 2 for 51-75, 3 for 76-100, 4 for 101-125, 5 for 126-150, 6 for 151-200, 7 for 201-300, 8 for 301 and higher 264 APPENDIX C Table A: Correlation Matrix X68 N83 VB64 VB65 SQVB65 X68 1.0000 N83 -0.31925 1.0000 VB64 0.17302 -0.027828 1.0000 VB65 -0.14877 0.054435 -0.27856 1.0000 SQVB65 -0.15664 0.049585 -0.22700 0.99633 1.0000 Table B: Correlation Matrix X69 N83 VB56 VB57 SQVB57 X69 1.0000 N83 -0.16228 1.0000 VB56 0.14985 0.19142 1.0000 VB57 0.070332 -0.14986 -0.59424 1.0000 SQVB57 0.065791 -0.13929 -0.58203 0.99748 1.0000 265 CHAPTER 6 CONCLUSION 6.1 Contributions of this Dissertation The need for the study of the role of social control in affecting human behavior is now generally recognized by economists. There has been, however, no attempt to formalize the effects of control on microeconomic behavior. Chapter 3 is an attempt to fill this lacuna. It may be recalled that a choice- theoretic model of individual utility-maximization in the presence of social control was assumed. It was demonstrated that, while such controls affect behavior, they do not determine it. The recursive nature of the model also helps us understand the mechanism through which the controls themselves change over time. It is important to note that neglect of the study of the process of change of control would lead to mere functionalism whereby any observed behavior may be attributable to the presence of some norm. Chapter 2 provides a summary of the literature on | social control from the social and biological sciences | 1 that should be of use to scholars of different fields. i i i i - . -..— 266 This is a significant step in the direction of bridging the gap between different disciplines in their approaches to common or interrelated problems. The complexity of human behavior in its diverse facets can be best understood through a multidisciplinary study and compartmentalization of knowledge into overly narrow fields is not helpful. An example is the recent literature on public choice theory which recognizes that human beings do not behave differently in the private and public sectors. In all endeavors, the goal of the self-interested decision-maker is to maximize satisfaction subject to whatever contraints (s)he faces. It is thus not plausible to model a person working for the government as an altruist concerned with the welfare of the people while one working for the private sector as an egotist. Public choice theory— which applies the methodology of economics to the study of political science— provides a realistic framework to understand the motives of actions of politicians and bureaucrats. Similarly, neoclassical economics usually portrays a "typical" individual as being devoid of culture, history, religion, beliefs. On the other hand, the traditional sociological 267 literature pictures a robot-like individual whose actions are largely, if not completely, determined by social norms. This dissertation emphasizes the necessity of combining the two: while individuals are affected by the social context, they are not mere robots. A further contribution is the empirical attempt to test the effect of mobility on social control. As discussed in Chapter 5, the data problems are formidable. There are difficulties in identifying and measuring variables that measure phenomena like mobility and control. Nevertheless, a dataset was found that enabled us to test quantitatively the effect of mobility on control, albeit in probabilistic terms. While the theoretical underpinnings of the reason for the negative relationship is rather common, there has been no effort until now to test this hypothesis empirically. Several other issues, however, need to be addressed. One issue that arises from this research j concerns the replicability of social institutions. In ! the following section, it is shown that mere mimicking i [ of institutions without careful analysis of the social 268 milieu may be disastrous. This may happen even when there appears to be very little difference between two societies. 6.2 The Problem of Replicability The importance of institutional factors and belief systems in the implementation of short- and long-term policy needs to be stressed. Furthermore, the need for examining not merely what may be termed economic variables such as income and consumption but also sociological and anthropological variables like norms and beliefs cannot be over-emphasized. Although the latter are more difficult to quantify, they are not necessarily less important than the former. The next several sections will demonstrate the problems that arise in trying to replicate institutions if proper attention is not given to the differences in | norms and beliefs from one society to another. I | Replicability, in this context, refers to the ease with j which a successful institution or organization can be | copied from one country to another, while retaining its i I original degree of success. Given the success of the Grameen Bank (GB) as described in Chapter 4, it is of 269 profound interest and importance to determine whether this experiment can be copied by other countries. A mere mimicking of institutions across countries will not succeed, unless careful attention is paid to the causal factors which led to the original success. The latter will often entail detailed consideration of sociological and anthropological factors in addition to the purely economic ones.1 6*3 Proiek Ikhtiar and GB: The Important Differences It will be shown that when the GB experiment was copied almost to its last detail in Malayasia, the initial result was not comparable to the resounding success in Bangladesh. This is in spite of the fact that both countries are predominantly Muslim and, at j least on the surface, appear to share many norms and ! beliefs. But— as critical mass theory would suggest— I I i small differences in individual dispositions can lead j 1 As set out in the model in Chapter 3, the utility ■ function of an individual contains arguments pertaining to both monetary and psychic components. To understand the latter component would require consideration of , sociological and anthropological factors. 270 to large differences in group behavior. It was inattention to these minor differences which, it will be argued, led to the initial failure in Malaysia. Projek IKHTIAR (or project to help the very poor strive to overcome their poverty) was a 2.5-year pilot project to test whether the GB experience could be helpful for the reduction of extreme rural poverty in the Northwest Selangor area of the Malaysian peninsula. Initially, there were several significant differences from the institutional setup of the GB. There were two specific targets: (i) disburse loans to at least 300 very poor households and (ii) maintain a repayment rate of at least 90%. The differences from GB included the following: (1) Projek IKHTIAR (PI) was not a bank nor was it connected organizationally with any bank. (2) PI obtained its loan capital from the Malaysian Islamic Economic Development Fund (YPEIM), which was concerned about "riba" or interest (usury?). There was no interest charged but, based on the concept 271 of "bai1 al-mudarabah" (or cost plus),2 a fixed administrative charge of about $30 per loan was charged. It will be recalled that GB charged interest at 16%, the same rate as the other banks in Bangladesh. (3) PI would write off any unpaid loan in the event of death of a borrower and did not institute an Emergency Fund. (4) Physical exercises at center meetings and saluting were deemed culturally inappropriate and were not adopted. (5) Since only about 25% of the households were in the target group (compared to close to 100% in the much 2This is a way (some would say a ruse) to avoid interest payments and thus observe the letter of the Islamic injunction against riba (or interest?). It is clear, however, that if the same fixed administrative charge is levied on any loan, those taking out smaller loans pay proportionately more. If, on the other hand, the charge is a percentage of the amount borrowed, then how is this different from interest, except perhaps in name? There is a further controversy not discussed in detail here because it is not directly relevant to the issues we are studying. There is substantial disagreement among Islamic scholars whether the Quranic ban is on any form of interest or on the specific institution of riba prevalent in seventh century Arabia. Riba in those days was a system whereby if a loanee failed to repay on time, the loan amount was doubled and redoubled; this had the almost certain effect of rendering him/her destitute and led to concentation of property in a few hands. 272 i | poorer rural Bangladesh), a rigorous "means test" was administered to identify the eligible villagers. (6) Although PI did not have an Emergency Fund, it did give out emergency loans from the general loan fund to reduce members' dependence on moneylenders. (7) Unlike the GB, PI is computerized down to the branch level, facilitating monitoring of loan repayment and staff performance. As Gibbons and Kasim [1990] point out, PI achieved success in: (a) reaching the target group, (b) increasing the income of borrowers, (c) diversifying income sources, (d) having an excellent repayment rate (95%) by women (although the overall repayment rate of 78% was disappointing), (e) increasing the savings of borrowers. There were, however, major problems and mistakes and this what we now turn to. I 6.4 The Mistakes I i The first mistake was the involvement of the i I j project staff in the process of group formation. Instead of letting villagers form their own groups, the | staff was "impatient and decided that [it] should assist by suggesting names and convening meetings." 273 Out of the first list of 110 qualified potential borrowers, only 2 groups were actually formed. These groups may be called "assisted groups" because, unlike the GB example, they were not self-selective. Because they were imposed from above by the overly zealous PI staff, they lacked the both the cohesiveness of the GB groups and the latters' sense of collective responsibility. It may be recalled that group formation is the cornerstone of the GB. The process of self-selection helps ensure having other members who are known and can be trusted and, perhaps most importantly, who are subject to group control in the form of sanctions in the event of non-compliance. It was quite understandable why the villagers were initially reluctant to undertake projects. The very poor have little margin for error and do not want to incur debts unless they are sure of being able to t ! repay. Moreover, as good Muslims, the elderly did not r ► perhaps want to risk dying with an unpaid loan. The I I proper institutional response in such situations (as ! | the GB success demonstrates) is painstaking | motivational work on the field which consists of overcoming both these fears. It appears that the PI 274 staff were interested in attempting to speed up the process of group formation in order to impress the donor agencies. The second problem was the prevalence of single purpose cultivation loans which the GB almost never makes. This introduces the problem of making weekly repayments prior to harvest. As defaults mounted and as there was no legal action taken, these borrowers stopped attending center meetings and repaying altogether. In the GB, there is a three-stage staggered disbursement procedure. The two poorest members receive loans first, then the next two, and finally the chairman. PI tried a two-stage policy whereby if the first two repaid regularly for 6 weeks, the remaining three (including the chairman) could obtain larger loans. It was felt that the creditworthiness of the group was established by the first two. But, as Gibbons and Kasim [1990] point out, this violated two basic principles: (1) adequate evidence of creditworthiness and (2) small and manageable first loans. Some members waited to borrow double the amount ! in the second round and then abscond. What the three- 275 stage GB method helps ensure is the kind of cohesion and social intercourse that is indispensable. In the initial stages, early repayment was allowed. This was a mistake becuase it "violated the staggered disbursement cycle and precluded the development of [g]roup feeling." The weekly repayment system implies that the members interact with each other on a regular basis and thus enhances cooperation. An additional problem was the practice of giving much larger loans if an earlier loan was repaid. Some people borrowed from friends (or moneylenders), paid off their first loans, obtained a large second loan, and absconded. The demonstration effect resulted in default of alarming proportions. Some members were taking out three loans simultaneously; not only were they taking out a project loan but they were also taking loans from the Group Fund and emergency loans from the general fund. The combined repayment of three separate loans constituted a burden too heavy for them to shoulder. Attendance at center meetings was irregular and/or tardy and the PI staff did not take a firm stand in this matter. It may be mentioned that, as in the case 276 of the GB, members pledge regular attendance at these meetings. Most group chairmen did not expend much effort to ensure members’ presence at these meetings. Additional problems included: (a) utilization of loan by husband (b) too many male borrowers and (c) mixing aid with credit, with the resultant confusion among receipients and borrowers. With regard to (a), as discussed in Chapter 4, the GB women members actively carry out business or trade and the loan taken out is rarely misappropriated by the husband. The position of the GB with respect to (b) is clear. Yunus [19 86] wrote, "GB attempted to reach the poor women ... from its very start." The reasons he cites for this include: (i) the fact that women are discriminated against by banks; (ii) they experience poverty and hunger most intensely; and (iii) they have the positive qualities required for development such as the drive to improve their lot, willingness to work hard, concern about their children's future, and the willingness to make short-run sacrifices to attain long-run benefits. This tilt toward women is borne out by the data: according to a recent newsletter published by the 277 Grameen Trust, about 93% of all GB members were women3. This compares to about 47%4 in the PI. With regard to (c), as discussed in Chapter 4, the GB borrowers pay the same interest as those of other banks in Bangladesh. The PI, on the other hand, charges no interest and may be viewed as a welfare scheme. Our purpose, however, is not to evaluate the success or failure of PI but to understand the role of social control. The evidence suggests that, but for the high repayment rate of women, PI would have been a total failure. The women borrowers are thus referred to as "penyelamat" (or saviors) of PI. The issue that needs to be addressed is: how is it that when a scheme that is universally hailed as a great success (GB) is replicated, with some (apparently^ minor changes, in a somewhat different (but similar) society, the same success is not observed? It is useful to remember that both Bangladesh and Malaysia are predominantly Muslim i j 3As of May 1992, there were 1,198,849 members of j whom 1,115,782 were female and 83,067 were male, j according to Grameen Dialogue. July 1992. Newsletter No. 11, Grameen Trust, Dhaka, Bangladesh. I Acomputed from data in Gibbons and Kasim [1990] 278 countries5 and that the members in both countries were poor villagers, although the extent of poverty is much greater in Bangladesh. One would expect, prima facie, a similarity of norms and beliefs. Furthermore, the loans in Malaysia were interest-free and this would, ceteris paribus. make it easier for borrowers to repay. As an added incentive for poorer borrowers, the fixed administrative charges on the first and second loans were subsidized at $10 and $2 0 respectively instead of $3 0. GB members, it may be recalled, paid interest at the rate of 16%. 6.5 Reasons for PI Failure At this point, PI sought the help of senior GB officials. The latter gave the former the following advice: "Get tough. Insist on strict discipline in [c]enter meetings and that all rules [are] followed, 5The one difference that is significant is that Bangladesh is ethnically almost homogeneous while Malaysia has ethnic diversity with Malay and Chinese populations. According to Worldmark Encyclopedia [1984], 96.5% of the population of Bangladesh is Bengali, predominantly a mixture of Aryan and Dravidian strains. The Malaysian population consists of 53.9% ! Malays and other indigenous groups, 34.9% Chinese and j 10.5% people who trace their origins from the Indian subcontinent. including [the one] that fellow [g]roup and [c]enter members take collective responsibility for anv absconders and/or arrears. Although it will not be easy for you and your staff and will be painful for some ... borrowers, you will be helping them in the long run.1,6 The problem was a subtle, yet significant, difference in Malay and Bengali culture. The former, Gibbons and Kasim [1990] point out, emphasizes "harmony as an ideal and tolerance of social deviants in practice." Strict observance of rules is reserved for religion and royalty; in other spheres of action, surface social harmony is of paramount importance. This implies tolerance of failure to perform even previously agreed-upon duties so that (surface) harmony is not disturbed. A further reason is religious (or perhaps fatalistic): since everything depends on the Wishes and Ways of Allah, (helpless) mortals will have to wait until He rectifies matters on Judgement Day. Thus, enforcement of agreements by appropriate sanctions on defaulters was almost nonexistent. 6emphasis added 280 In the absence of effective collective responsibility, it is not surprising that PI was encountering severe problems. As Yunus quoted in Gibbons and Kasim [1990] put it, "Formation of the small five-member group of the members' own choosing, and federating the groups into centers, helps create the right kind of peer pressure ... when a member tries ... to violate GB rules and peer support ... when a member falls into any difficulty."7 In the absence of informal norms to facilitate such collective responsibility (what Coleman [1988] would call social capital), a proper institutional response was called for. In other words, the social capital needed to be created formally. The norms did not permit informal enforcement of strict discipline and rigorous performance of duties. Deviations could be tolerated as long as the "general direction" was correct. Refusal to accept deviation I makes one responsible for causing social disharmony. i The norm in Malay society is that one should not ; 7The (incomplete) citation given is Yunus, M. The ; Grameen Reader, published by David S. Gibbons. It was I not possible to locate this work. 281 intervene in a neighbor's affairs even if (s)he was breaking his/her solemn promise and if others would be hurt by such non-compliance.8 In other words, collective responsibility could not be attained in Malaysia without some formal institutional mechanism. 6.6 The Institutional Response The Special Center Savings Fund (SSF) was set up by PI to reguire group and center members to institutionalize collective responsibility. It was triggered automatically if any default was not removed by the subsequent center meeting. Each member of the group in which the defaulter was a member had to pay an additional $2 per week into the SSF and the other center members had to pay $1 per week until the default was covered. The members were initially hostile to the idea of the SSF. Its implementation had the following overall result: the immediate effect was a further sharp rise 8This may be contrasted with the example cited in : Chapter 4 where forty GB women members threatened the husband of a female GB member with dire consequences unless he consented to his wife's continued membership in the GB. 282 in the number of defaulters (primarily due to refusal to pay into the SSF). If a group had a defaulter, it had to pay $2 per week even when additional members fell into arrears. The duration of the weekly payment would increase but there was no immediate increase in the amount of payment as the number of defaulters increased. In the same way, members of other groups in the center had to pay $1 per week irrespective of how many groups were in arrears. The SSF did not help solve the existing repayment problem and actually made it worse in groups and centers with default. What it did serve to do was to stop repayment problems from spreading to centers without arrear problems and to new centers being established. Figures 1 and 29 represent typical examples of a center with existing default and one with none. In rural Bangladesh, the elected center chief goes to the defaulter's house and discusses the problem directly with him/her. If necessary, (s)he seizes the goods bought with the loan and sell them to recover 9See Appendix. 283 part or all of the loan. Husbands who misuse their wives' loans are visited by the entire center and ridiculed/cajoled/threatened to pay up.10 The norms and cultural context there make these types of behavior acceptable. In Malaysia, the SSF is an alternative of ensuring collective responsibility, in the absence of which GB-type schemes cannot succeed. Careful attention to subtle differences in norms and beliefs helped rescue the PI. It is of vital importance to consider such differences and make appropriate instututional responses. Thus, close attention needs to be paid to sociological and anthropological data, collection of which is expensive in terms of money and time. In the case examined above, a superficial (and incorrect) analysis may well conclude that the norms in rural Bangladesh and rural Malaysia would be similar given that the overwhelming majority of both populations is Muslim. Further study would demonstrate that similarity in religion does not imply coincidence of all norms. On the other hand, it is perhaps safe to speculate that similarity of 10See footnote 8 (above) for an example. 284 religion may imply coincidence of many (even most) norms. But, in group behavior, as critical mass theory would suggest, even small differences in norms would have large differences in group outcome. 6.7 Three Other Examples Briefly Considered Published material in professional journals or books about examples of GB replicability in other countries is hard to obtain. Recourse is thus taken to journal and newspaper articles to fill this void. 6.7.1 Mali. Honduras, and Nepal The Economist of August 20, 1988— in an article entitled "Patching up poverty"— discusses the role of the International Fund for Agricultural Development (IFAD). It singles out the GB as an example of how banks can give loans to small groups of people without collateral and achieve repayment rates of 99%. The GB experiment, with some variations, was tried in Mali. A local bank offers to lend money to villagers if they contribute toward a village credit fund which would constitute 10% of the start-up money. The original capital is turned into a revolving fund and monitored by the villagers. IFAD claims that the 285 project, which started in 1984, has a repayment rate of 100%. No further information was available on the Mali experiment. It is not known, for example, whether and how groups were formed. But, like the GB, the emphasis is on self-sufficiency and not reliance on government handouts. The initial 10% contribution can be construed to be an expression of earnestness and sincerity of purpose. In Honduras, credit was given to small groups manufacturing cheese, paper and brooms, rearing pigs, and processing henequen. The repayment rate there is 85% to 90%. A similar scheme in Nepal is geared toward female heads of households who are landless and unemployed. Loans are given for handlooms, for weaving and making bamboo furniture. The repayment rate for the Nepalese scheme is not given but the article implies that it too is doing well. 6.7.2 Bolivia The Wall Street Journal of April 10, 1992— in an article entitled "Profit's Not a Dirty Word at Bolivia's Bank for the Poor"— discusses the 286 accomplishments and future prospects of Bolivia's Micro Enterprise Promotion and Development Foundation (Prodem). Prodem's clients— 70% of whom are women— have a repayment rate of well over 99%. These loans were given on the basis of personal guarantees to desperately poor street vendors and tradespeople. The article goes on to say that the "effective interest rate" was 45%. In the absence of any further information, one can only surmise what this means. It will be recalled that, in Chapter 4, the effective rate of interest on GB loans was computed at 3 0%. This rate was arrived at by adding the compulsory contributions to the Group Fund and the Emergency Fund to the explicit interest charged (16%). It seems safe to speculate that some form of forced savings along the lines of the GB is a part of the institutional setup of Prodem. j 6.7.3 The U.S. J The Wall Street Journal of June 23, 1992 j summarizes the results of trying to replicate the GB in , the U.S. The Women's Self-Employment Project (WSEP), 287 launched in Chicago in 1986, sticks closely to the GB model. Five members form a "borrowing circle", although it is not clear from the article whether the circle is self-selective. The five members choose two to get he first loans and, if these two repay regularly for six weeks and if all five members attend three consecutive circle meetings (held twice a month), the third member is eligible. The repayment rate has been 100%. The Good Faith Fund in rural Arkansas did not find peer pressure sufficient. In the first two years of operation, the default rate was 40%. It moved away from the classic GB model. The reason given by the director is, "Peer pressure isn't as significant as it might be in a place like Bangladesh." Note the apparent anomaly: one would, a priori, expect peer pressure in rural Arkansas to be more similar to rural Bangladesh than would urban Chicago. How then does one i ( exlain the success in Chicago and the failure xn | Arkansas? To do this would require data about both j these cases which are not available. As was shown in ] the Malaysian example above, even a small difference in i disposition may lead to a big difference in outcome. 288 To the extent that data are available, it seems safe to conclude that the Chicago scheme pays more attention to the development of group feelings and camaderie with its insistence on regular circle meetings and this alone may well be a primary contributing factor to its success. This provides support for the hypothesis that a properly designed institution can enhance group feelings (as in the GB example) or perhaps even develop them (as in the Chicago example). 6.8 Control and Freedom It is important to address the question of freedom versus control at this stage. Does the prevalence of control neccessarilv imply reduced freedom? It may be recalled that, in Chapter 4, a case was described where forty women marched on to the home of a man and threatened him with dire consequences if he continued to object to the membership of his wife in the GB. Was this man's freedom thus curtailed? At the very outset, it is clear that the man by attempting to impose his views on his wife was ! restricting her freedom of choice. In this example, l ! the question is essentially this: should women have the 289 right to choose their occupation freely? If the answer to this question is in the affirmative, then the actions of the forty women have furthered the cause of freedom and not hurt it. One's freedom does not include the power to restrict that of others. Following Sen [1988], the freedom to choose is a particularly important functioning.11 The actual freedom to choose is sometimes limited even though there may be little or no restraints on freedom. Sen criticizes the libertarian preoccupation with the latter and notes that development economics should judge a society in terms of the actual freedom achieved rather than by the processes through which this achievement came about. He does not, however, deny the importance of the positive freedom to choose, given the level of other functionings. Sen defines a person's capability as the set of alternative n-tuples any one of which the person can choose. Well-being depends not only on the actual functioning achieved but also on the capability set 11Functioning refers to the ability to do certain I things and to achieve certain types of beings, such as, I being well nourished, being free from avoidable : morbidity, being able to move freely. 290 from which that n-tuple of functionings is chosen. Sen [1985] defines "refined functioning", x/S, when one chooses to do x out of any member of a set S. This enables us to distinguish between fasting when one has the choice of not starving and willy-nilly starving. While there may be agreement as to which functionings are valuable, there may well be disagreement on their relative values. When we arrive at an agreed list, the principle of "dominance"12 identifies a minimal partial order on the evaluation space. In order to extend this partial order, one has to narrow the range of the weights. In the GB example, conjugal privacy and the right of a wife to choose her occupation may both be considered functionings. The principle of dominance may allow us to choose between less of one and more of the other, depending on the weights or relative values attached to each. In the event that this does not provide a clear answer, the 1 range of the weights would have to be narrowed to reach I | an unambiguous result. While such exercises may be i I extremely difficult, they need nonetheless to be i - - - - - - - - - - - - - - - - - - - - - - - - - - 12that is, greater achievement in every respect. 291 undertaken in order to assess the effects of social change. It seems reasonable to conclude that social control of the type that expands the choice set contributes to human freedom in the sense that the number of functionings available increases. As discussed in Chapter 4, the ties that bind together the rural Bangladeshi community make possible the success of the GB. If one were to examine superficially only the sanctions imposed— for example, the forty angry women marching on the man's house or the social boycott of a defaulter— one would err in not recognizing the potential freedom-enhancing role of these very same sanctions. This is not to deny that social controls may not under certain circumstances be contradictory to the goals of freedom, such as in the case of the caste system. What is required is a careful and systemmatic study of each situation. There are two further problems in the evaluation I of development as enunciated in Sen [1988]: (1) the i i valuation functions accepted by different people may i i j differ ("value-heterogeneity"); and (2) the process of j development may change people's valuations ("value- 292 endogeneity"). It is important to recognize that values may differ not only within a given society but also between societies.13 Thus, the right to privacy may be valued more in Western societies than in many Asian societies. On the other hand, in many Asian societies it may be considered an obligation to help a poor relative while such an obligation may not exist in Western societies. Can we conclude that one is better (or more free) than the other? Once again, the principle of dominance and narrowing of weights may help answer questions such as these. The process of observation may involve biases and may entail a trial- and-error methodology. It is helpful to remember that the study of complex social processes is extremely difficult and it is not likely that we shall obtain straightforward answers. It is also important to recognize that there is no reason to expect that the | orderings of the functionings will be the same in all I ] societies. Failure to do so may result in the i 13Sen [1988] considers only the first type of ! value-heterogenity, although it seems safe to infer I that he would agree that the second type exists also. 293 unwarranted imposition of values by one society on an unwilling (or even hostile) one. Value-endogeneity is an even more difficult problem; the valuation function itself may depend on what is being valued. The process of development may change perceptions regarding what is valuable and what weights are attached to it. If there is a shift from considering profits as immoral, it is almost certain that trading activities will be viewed more favorably than before. In Chapter 4, the new role of female GB members as harbingers of future change was thus emphasized. In a traditional rural Muslim society, the mere fact that women are taking more control of their lives is a profound shift away from the past. Thus, while it may have been the case that the valuations ordering of the past would have given preference to housewives rather than businesswomen, that no longer seems to be the case. Such changes are endogeneous in the sense that the process of development itself causes these changes in attitudes and norms. 6.9 Recommendations and Conclusions 294 The major recommendation of this dissertation is that economists need to remember that they are social scientists. The human being is a social animal and is influenced (though not completely controlled) by social norms and beliefs. It is thus crucial that "economic" studies should try— within their respective budget constraints— to collect relevant (subjective) data about people's conceptions with regard to ethics, morality and religion. An anthropological study involving a fairly long stay in the survey area by the researcher would be ideal. Failing this, however, at the minimum, efforts should be made to understand such phenomena through the standard questionnaire method. The Indian survey discussed in Chapter 5 was the only one we found that could be used to study the effects of norms on human behavior and for this we are thankful to the researchers who carried out that survey. Too often, one gets the impression that human beings are machines with no known history, culture and religion and this sad state of affairs has to end. 295 Bardhan [1989]14 is an interesting work that endeavors to bridge the methodological gap between economists and anthropologists in the measurement of economic change. Bardhan [1989] contrasts "the image of economists ... [as] "mimic physicists" crackling with the static of electronic data crunching while anthropologists smell of the outdoors and manure." Vaidyanathan [1989] points out that, while many socio economic phenomena are quantifiable, some are not. Thus, a proper understanding of such phenomena requires a combination of both the macro and micro approaches.15 According to Bardhan [1989]16, the macro surveys are particularly weak in capturing "dynamics, processes and relations". Even in the rare 14This citation and the next refer to the edited book and not the introductory section written by Bardhan. 15The macro approach— generally favored by economists— consists of large-scale surveys designed to obtain information on certain characteristics of the entire population (censuses, for example) or from a representative cross-section thereof (sample surveys). The micro approach— favored by anthropologists— refers to detailed investigations of specific localities and groups, by a participant-observer unhindered by J presuppositions of what is important. 16This citation is to his introductory article. 296 large-scale longitudinal surveys, the method is one of comparative statics, not disequilibrium dynamics or adjustment paths. Only intensive case studies can detect "telltale straws in the wind in fluid transitional situations". Such changes may appear statistically insignificant but may be harbingers of substantial future change, and may be understood only by a perceptive observer.17 As noted in Section 5.3.3, studying the existence and effect of social norms is a complex, though essential, undertaking. Such an effort would ideally involve living in the society one wishes to investigate for an extended period of time. Quantitative data are supplemented by qualitative data regarding norms, religion, culture and belief systems. Perhaps equally importantly, subjective information regarding norms should be compared to actual behavior. People may state that the caste system is immoral; the empirical question is: does their behavior reflect their avowed distaste? Time and monetary constraints did not allow 17It is the opinion of this author that the I changes in the attitude of women members of the Grameen ; Bank are precisely of this type. For more details, see Section 4.5.3. 297 us to undertake such a detailed study. Reliance was made on observations of other researchers and thus the possibility of bias exists. It should nevertheless be emphasized that since much of the empirical work on norms were in the context of Bangladesh, the fact that the author was born and raised in that country and is well aware of its social milieu implies that the probability of such bias is low. Often, economists have the tendency to view norms as "constraints" that shrink the choice set. It was the purpose of this dissertation to demonstrate that such a view is not always justified. As Coleman [1988] has shown, norms can play the role of "social capital" that facilitates mutually beneficial exchange which would not take place in their absence. This can be elaborated by considering the pioneering article by ! j Stiglitz and Weiss [1981] regarding credit and information. The above article seeks to explain why the textbook "quantity demanded (qd) equals quantity supplied (qs) equilibrium" may not be relevant in a J credit market with imperfect information. At some | interest rate r*, qd may be greater than qs but there 298 may be no tendency for the price (that is, r) to rise above r*. This is because the bank may judge loans at r > r* to be more risky than the average loan at r* and thus the expected return to a loan at r > r* may be lower than that to one at r*. The imperfect information that the bank has— specifically, in not being able to distinguish between risky and safe borrowers— restricts the size of the market. In the GB case, as noted above, such information scarcity is mitigated by the self-selective group.18 A further interesting and largely overlooked point in the same article also merits discussion. The formal banking sector relies on collateral as security for loans they make. In situations where qd > qs, it may appear that the banks should increase collateral requirements and thus reduce qd. But this may increase I risk for two reasons. If smaller projects are more likely to fail and if all potential borrowers have the 18As discussed in Chapter 5, interlinkage may also serve to facilitate the flow of information. Thus, a landlord may be willing to lend to his/her tenant because the former has more knowledge about the latter than another potential borrower about whom little is known. The tenant with whom the landlord has a i multiplex relationship would be considered less risky ceteris paribus. 299 same amount of equity, increasing collateral will imply financing smaller and, thus, riskier, projects.19 Even if potential borrowers have different equity, a similar adverse selection problem may arise. The wealthy borrowers may be those who were lucky at past risky ventures. They are then likely to be less risk averse than the other group and may pose a greater risk to the bank even though they may be willing to furnish more collateral. Collateral serves as a signalling device where the bank is unable to evaluate, and exercise control over, individual loanees. In a scheme like the GB, evaluation of creditworthiness, monitoring loanees, and sanctioning violaters are done at a much more personal level and thus collateral can be 19Note the relevance of this to the GB case where the potential borrowers are similar and it is quite j likely that the very smallest projects may be riskier 1 (especially if one remembers that the average project I is small and that the repayments are weekly). The importance of zero collateral is that any available resource may go toward increasing the project size and, thus, reducing its risk. Viewed from another angle then, the amount of collateral is the cost that has to be borne when the "social capital" in the form of | information is not available or is available in ; insufficient quantity. 300 dispensed with, thereby enlarging the scope of credit transactions20. Arnott and Stiglitz [1991] examine the situation where insurance is characterized by moral hazard. It is obvious that moral hazard causes fully insured people to expend too little effort.21 The insurance firm is unable to monitor the actions of the insured and reacts by rationing the amount of insurance at a given price. Individuals resort to nonmarket insurance, that is, supplememtary mutual assistance between family and friends which is not observable to the insurance firm. It is clear that the family and friends are ex- ante in a better position to monitor the insured. The issue of crucial importance is whether such capability can be actualized. If this monitoring does not occur, nonmarket insurance lowers welfare because it leads 20The signal of creditworthiness in the GB is not the amount of collateral (GB members require rig collateral) but the viability of the project, the seriousness of purpose of the loanee, and the reliability of the loanee (reflected in her/his ability i to be self-selected in a group). 21 If one's car is fully insured, for example, one is less likely to install an alarm system. 301 individuals to take less care. If, however, effective monitoring occurs, the moral hazard problem is mitigated. There are then two offsetting effects: (a) the risk-pooling advantage of the market insurance and (b) the monitoring advantage of nonmarket insurance. With perfect monitoring, the latter is shown to predominate and nonmarket insurance enhances welfare. The monitoring can take three forms: (i) direct monitoring where the principal monitors the agent herself/himself; (ii) supervision where a supervisor is hired to monitor agents; and (iii) indirect monitoring where a mechanism is set up to induce agents to monitor themselves. The literature has largely ignored the third form of monitoring which will encourage peer monitoring through the recognition of the interdependence of agents' utilities. In the GB example, this is precisely what is at work. In fact, | the GB has gone one step further and has narrowed, if not eliminated, the distinction between principal and agent. A person found eligible for credit is not a loanee, but a member and a shareholder in the bank. I Furthermore, as discussed in Chapter 4, loan i applications are approved by the center chief and such 302 decisions are almost always accepted by the GB officials. What can we say about the likelihood that monitoring will actually take place? In addition to the observability of effort, Arnott and Stiglitz [1991] identify factors such as the duration of the partnership, the severity of punishment for default, the power of reputation and social pressure, and nonmonetary rewards. In the GB, the process of self selection would usually imply groups whose members are known to each other for long periods of time. The collective sanctions imposed in the case of default make such actions costly.22 As discussed in Chapter 4, social pressure in the form of enforceable norms is strong in traditional (and relatively immobile) societies and was depicted in our model by the Ci term. Finally, the nonmonetary or psychic rewards— the c^. term in our model— is important, though often overlooked. A person may engage in an activity even in I 22It may be recalled that if one member defaults, | the entire group is denied access to credit. 303 the absence of monetary reward if (s)he believes that it is the right thing to do.23 It seems reasonable to conclude that the GB has both the advantages of risk-pooling and better monitoring. With 961 branches and 1,198,849 members24 spread throughout the country, there is a sufficiently large number of (geographically-dispersed) people to provide effective risk-pooling.25 At the same time, the small primary group size (of five members) and the self-selecting process of group formation has mitigated the problem of moral hazard. As discussed in Chapter 4, social ostracism and direct intervention have reduced default to almost zero. What is observed is a harmonious blend of a formal institution with the informal milieu pertaining in the country. The GB acts as an institution that takes advantage of the existing , 23For details, see Chapter 3. j 24as of May 1992. This information was obtained I from: Grameen Dialogue. July 1992. Newsletter No. 11, ! Grameen Trust, Dhaka, Bangladesh. 25Thus, in the event of a natural calamity, of which Bangladesh unfortunately seems to get a disproportionate share, the risk-pooling works except ' in the unlikely event that the entire country is • affected. 304 social conditions and builds on it, without attempting to impose solutions from above.26 It recognizes that the individuals have more information about each other and that a bank can never be a substitute for the proper dissemination of this information; all it does is serve as a conduit for the steady flow of such information. The norm of caring for one's elderly parents is of vital importance in societies where formal pensions schemes are rare. Before undertaking policy initiatives, careful attention should be paid to the existence and usefulness of this norm. A certain society may allow the father to will his property to any offspring he wants to. For the sake of "fairness" and "egalitarianism", some reformers may object to this and pass a law requiring the father to divide his property equally among his offspring. But what this law would result in is the father's inability to use bequest as a strategic bargaining chip to increase his chances of obtaining old-age support. In the absence I » j “which is unfortunately true for most credit j programs 305 of formal pension schemes, the informal ones should be carefully nurtured and promoted.27 Although this dissertation applied the concept of social control to help understand the success of the GB and the future of old-age support, other fields of inquiry can also benefit from the systemmatic study of norms. Family planning schemes are important government undertakings in many developing countries. The acceptability of such schemes is not likely to be the same in a Muslim or Catholic country compared to other countries due to perceived religious injunctions. In terms of our model,28 even if the monetary or tangible benefits of not having a child is greater than that of having one, the psychic costs may be different. If there is an internalized norm regarding the desirability of having many children and/or the undesirability of using contraceptives or having ; abortions, the person may choose to have more children even though the net effect may be to lower his/her monetary benefits. It is perhaps because of 27see Nugent [1990] for similar arguments i 28For details, see Section 3.5.3 306 internalized norms such as these that the exact cause and timing of the fertility revolution is difficult to ascertain. Thus, one implication of this dissertation is that the understanding of decisions pertaining to family size is facilitated by proper consideration of existing norms. It also needs emphasizing that different institutional responses may be required under different social conditions. As noted above, in Bangladesh individuals can help ensure cooperation by sanctioning violaters and there appears to be no need for a formal sanctioning institution at this point in time. In Malaysia, however, since such behavior is not acceptable, a formal mechanism to monitor and sanction defaulters is required. Only a careful study of the values, norms and belief systems of a particular society can help in the designing of successful institutional responses tailored to the needs of a unique society. 307 BIBLIOGRAPHY Arnott, Richard and Joseph E. Stiglitz. 1991. "Moral Hazard and Nonmarket Institutions: Dysfunctional Crowding Out or Peer Monitoring?", American Economic Review, Vol. 81, No. 1. Bardhan, Pranab. 1989. "Introduction" in Pranab Bardhan (ed.), Conversations Between Economists and Anthropologists. Oxford University Press, Delhi. Coleman, James S. 1988. "Social Capital in the Creation of Human Capital". American Journal of Sociology 94 Supplement: S95-12 0. Gibbons, David S. and Sukor Kasim. 1990. Banking on the Rural Poor in Peninsular Malaysia. Center for Policy Research, University Sains Malaysia, 11800 Penang Malaysia. Hossain, Mahabub. 1986. "Credit for the Alleviation of Rural Poverty: The Experience of Grameen Bank in Bangladesh", Bangladesh Institute of Development Studies, Dhaka and International Food Policy Research Institute, Washington, D.C. Nugent, Jeffrey B. 1990. "Old Age Security and the Defense of Social Norms", Journal of Cross- Cultural Gerontology. Vol. 5, No. 3. Sen, Amartya. 1985. "Well-being, agency and freedom: The Dewey lectures 1984", Journal of Philosophy, 82. - 1988. "The Concept of Development", in Hollis Chenery and T. N. Srinivasan (eds.), Handbook of Development Economics. Volume 1. North Holland, Amsterdam. Stiglitz, Joseph E. and Andrew Weiss. 1981. "Credit Rationing in Markets with Imperfect Information", The American Economic Review. 308 Vaidyanathan, A. 1989. "Macro and Micro Approaches to Studying Rural Economic Change: Some Pointers from Indian Experience", in Pranab Bardhan (ed.), Conversations Between Economists and Anthropologists. Oxford University Press, Delhi. Yunus, M. 1986. Grameen Bank: As I See It. International Labor Office, November 1986. Worldmark Encyclopedia of the Nations. 1984. Vol. 4, Worldmark Press, Ltd., New York. i Numoer of hemoers In the Center - 2b 309 APPENDIX PUSAT SINAR MURN! Pattern of Arrears, June 1987 - June 1988 SSF J87 J J8 8 F A S O D M A M J Weekly Center Meetings, June '87 — June '88 Figure 1 Source: Gibbons and Kasim [1990], pp 49. Note: SSF was started in September, 1987. Number o f Members I n the Center • 26 310 PUSAT SRI HIDAYAH Pattern of Arreas, June 1987 - June 1988 20 10 0 J 8 8 F J 87 J Weekly Center Meetings, June '87 — June '88 Figure 2 Source: Gibbons and Kasim [1990], pp 52. Note: SSF was started in September, 1987.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Weight variations among preschoolers: An analysis of evidence from rural Tamilnadu

PDF

The influence of industrial pressure groups within the People's Republic of China

PDF

Irrigation institutions and technical change: A case study from South India

PDF

Life-cycle demographic decisions and the economic well-being of single parents in the United States, 1964-1989

PDF

The economics of dowry

PDF

Patronage, dependence and debt in peasant agriculture

PDF

Countertrade: A theoretical and empirical investigation

PDF

Knowledge and use of the contraceptive effect of breastfeeding: Evidence from Malaysia and Guatemala

PDF

Adaptive economizing in disequilibrium: Essays on economic dynamics

PDF

Adaptation, assets, and aspiration. Three essays on the economics of subjective well-being

PDF

The size distriubition of Turkish manufacturing establishments across sectors and over time and its responsiveness to economic policy changes

PDF

A case study of underutilization of canal irrigation in Bihar, India

PDF

Voluntary overcompliance of environmental standards: Theory and empirical evidence

PDF

Dutch disease economics, rent-seeking theory and the Nigerian food sector

PDF

Moving toward free trade in North America

PDF

A reexamination of Hobbes' world: Power, property and markets

PDF

Money, credit and economic growth in fragmented economies

PDF

China's economic reforms and establishment size, revealed comparative advantage, and policy effects: A new institutional economics approach

PDF

The price of oil and OPEC behavior: A utility maximization model

PDF

Three essays on the determinants of firm structure and contractual choice

Asset Metadata

Creator Chowdhury, Mussaddequddin (author)

Core Title An economic analysis of cooperative group behavior: Theory and evidence

Degree Doctor of Philosophy

Degree Program Economics

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag Economics, General,OAI-PMH Harvest

Language English

Contributor Digitized by ProQuest (provenance)

Advisor Nugent, Jeffrey B. (committee chair), Easterlin, Richard A. (committee member), Kuran, Timur (committee member), Van Arsdol, Maurice D., Jr. (committee member)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c20-279636

Unique identifier UC11258957

Identifier DP23396.pdf (filename),usctheses-c20-279636 (legacy record id)

Legacy Identifier DP23396.pdf

Dmrecord 279636

Document Type Dissertation

Rights Chowdhury, Mussaddequddin

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA