COPYRIGHT Frank Cass & Company Ltd. (UK) 1994

1984: 1289^ -- and the economist Field -- for whom 'rules logically antedate the market' |Field, 1981: 193; also 1984^ -- have their thinking anchored in the other realm where no spontaneous order exists. Recently game theorists have shown interest in the question of the emergence of rules or norms of cooperation. A well-known group of them have adopted the (Hayekian) evolutionary perspective according to which rules and (moral) beliefs of a free society are the unintended outcome of a process of evolution occurring in the absence of any conscious human design. Thus, for Sugden, the conventions, or established patterns of behaviour, which create order in a market society are supported by moral beliefs (people believe that they ought to keep to these conventions). Yet 'there is no independent principle of justice that provides a rational basis for these beliefs' because 'the

AUTHOR ABSTRACT: This two-part article is an attempt to clarify the social conditions upon which the viability and efficiency of the market system rest. It strives to show that the 'embeddedness' thesis, that is, an explanation based upon the existence of long-run personal ties involving the use of reputation mechanisms among transactors, cannot fully elucidate the question as to how the problem of trust is solved in market societies. As explained in Part I, there are difficulties of both theoretical and empirical/historical kinds and these explain why the 'market order' needs to be sustained by private and public order institutions. In Part II, the role of generalised morality in backing or supplementing such institutions is discussed in the light of game theory, and particular emphasis is put on the ability of moral norms to sustain honest behaviour by generating the right kind of preferences and establishing trust. The vexed problem of the dynamics of norm emergence and erosion is then addressed with a view to showing that norms of generalised morality -- perhaps contrary to moral norms in small groups -- cannot be easily created by fiat nor be expected to evolve spontaneously when they are needed to make economic exchanges viable. Ultimately, the cultural endowment of a society plays a determining role in shaping its specific growth trajectory, and history therefore matters. Finally, to illustrate the theoretical discourse, reference is made to present-day Third World countries. It is argued that economic development is especially difficult in countries where norms of limited-group morality prevail and do not readily give way to generalised morality.

II. TRUST AND GENERALISED MORALITY

The Problem of Political Order

The argument presented in Part I of this article has led to the conclusion that dense networks of close and continuous business relationships are not sufficient to solve the problem of trust in market exchanges when the division of labour is well developed. Private and public order institutions are needed to create order in the market. In particular, the state has a critical role to play which goes far beyond that of establishing or strengthening self-enforcing mechanisms for the control of fraud and deceit. Acknowledging this much, the problem is not solved, however. Indeed, if it is relatively easy for the state to solve coordination problems -- that is, problems for which self-enforcing rules or mechanisms can be designed whether they correspond to pure conventions (for example, the laying down of a measurement system) or to processes of information centralisation -- the same cannot obviously be said of those PD problems (problems having the form of a prisoner's dilemma) that can be surmounted only by means of sanction mechanisms.

As a matter of fact, under circumstances in which they know that each of them has an incentive unilaterally to violate rules which they otherwise support, individuals may be expected to demand strong sanction systems so as to prevent rule-breaking and thereby reassure everybody that the rules are well abided by. Particularly in situations where fraud and deceit are widespread, people may thus actively or passively support the emergence of a strong state, even an authoritarian state bent on restoring 'law and order'. As pointed out by Putnam with the case of Southern Italy in mind, it is ironically 'the amoral individualists of the less civic region who find themselves clamoring for sterner law enforcement' |Putnam, 1993: 112^. The obvious difficulty with this solution is that it can prove very costly for the society in terms of both material resources and loss of freedom. Further, it can be largely ineffective inasmuch as, due to imperfect information about the agents' particular actions, the state cannot avoid making errors in imposing punishments. As a consequence, free riders know they have a chance to escape punishment while honest people know they can suffer unjustified sanction. If errors are too important, agents may be incited to give in to fraud instead of refraining from it (for proof, see Ostrom |1990: 8-12^).(1) Clearly, the fundamental problem of the market order cannot be solved by simply referring to the existence of highly evolved political and juridical institutions |Elster, 1989a: 276^.

To economise on information costs, a (totalitarian) state may be tempted to rely on a self-enforcing mechanism that is especially perverse, namely denunciation of deviant behaviour by the people themselves (possibly against some attractive rewards, material or symbolic). This old strategy, which comes down to vesting prosecutorial powers in the public, has considerable drawbacks which make it self-defeating. In particular, it is bound to entail formidable indirect costs, at least in the long run, if only because the distrust thereby created unavoidably gives rise to severe excesses or injustices (all the more so as many dictators have paranoid personalities prompt to fabricate imaginary enemies) and spills over into all spheres of human interaction. Fear and suspicion infiltrate the whole society as a result of which individuals tend to withdraw from as many social intercourses (including economic transactions) as possible, to shun away from all risky ventures, and to take refuge in private life. An appalling illustration of the extremely tragic consequences that the above strategy can entail is provided by Stalinist USSR.(2) In the words of MacDonald |1991: 120-23; 220-21; emphasis added^:

Under Article 58, Section 12 of the Soviet Criminal Code (1926), failure to denounce anyone guilty of crimes listed in other sections of the Code was punishable by death or imprisonment for an unlimited period. As a corollary of this, the 'duty to inform' was endlessly played upon by the government and the security organs . . . the practice of denouncing neighbours and even relations out of sheer malice was common, while 'unmasking' one's workmates rapidly became an accepted way of gaining promotion . . . Under Khrushchev, the use of civilians to police civilians attained a new sophistication, notably in the form of the druzhiny or People's Guards, authorised by the Supreme Soviet in 1959 to rid the streets of 'parasites' and harass anyone generally failing to conform. By exhorting neighbours to spy on each other, sanctioning 'comrades' courts', and encouraging the Komsomol in their traditional role as moral vigilantes, the Soviet authorities mobilised 'the wrath of the People' against the people themselves . . . |Eventually^ social cohesion had completely collapsed. People no longer trusted each other and love was a rare and improbable bloom hanging on here and there in a landscape of sterile grey . . . everyone in Russia was in effect placed in solitary confinement.

In the case of the USSR (and Communist China), exploitation of people's distrust for the sake of making them compliant with the state's rules was actually combined with huge investments in legitimacy which were themselves costly. As a result of 'propagandist brainwashing' and generalised distrust, 'the word "conscience" had gone out of ordinary use since its function had been taken over first by "class feelings" and later by 'the good of the state"' |ibid. 123; also Lane, 1991: 217^. It would thus appear that, for the political order ruling over a market economy (which, of course, the USSR was not) to be viable and not too oppressive, two conditions must be satisfied. For one thing, the domain of public power must be restricted, implying that many aspects of social intercourse which create opportunities for conflict ought to be organised by the individuals themselves, according to certain informal standards or fundamental rules for mutual tolerance. As noted by Buchanan: 'Life in society, as we know it, would probably be intolerable if formal rules should be required for each and every area where interpersonal conflict might arise.' For another thing, to the extent that formal rules are needed to resolve conflicts, the costs of operating law-enforcing agencies must not be too high |Buchanan, 1975: 118-19; also Shott, 1979: 1329^. Arrow who has clearly perceived the same problem pointed to ethics as the only way to solve it in a satisfactory manner:

. . . ethical elements enter in some measure into every contract; without them, no market could function. There is an element of trust in every transaction; . . . It is not adequate to argue that there are enforcement mechanisms, such as police and the courts; these are themselves services bought and sold, and it has to be asked why they will in fact do what they have contracted to do (Arrow |1973: 24^ quoted from Williamson |1985: 405^); |also Arrow, 1971: 22; Phelps, 1975^.

In the following, it is argued that the pervasive presence of generalised morality in a society can prevent enforcement costs of the rules of honesty from being excessively high -- perhaps to the point of making the system unworkable |North, 1981: 45^ -- while avoiding recourse to perverse mechanisms that have the effect of enhancing distrust instead of establishing trust. In fact moral norms act as a substitute for, or a reinforcement of, state-engineered rules or control mechanisms, with the result that enforcement and punishment institutions become of secondary importance. The first function derives from the fact that individuals internalise public good considerations while the second one arises from their conviction that the law represents the public good. In this perspective, civic consciousness appears as this attitude of respect for the law which tends to prevail among citizens who have such a conviction. When individuals thus place a high value on obedience to law, the announcement and enactment of the rules that constrain behaviour suffice to cause compliance.(3) To have a sound analytical grasp of the precise role of generalised morality, it is useful to start by posing the problem of mutual trust as one of convergent expectations of the right kind.

Trust as a Matter of Convergent Expectations

A static view: Mutual trust among the people is evidently what a society needs to be able to sustain a market order on a long-term basis and at reasonable (transaction) costs. To probe into the matter further, let us first follow Dasgupta by defining trust as correct expectations about the actions of other people that have a bearing on one's own choice of action when that action must be chosen before one can monitor the actions of those others' |Dasgupta, 1988: 51^. To better figure out how or in what conditions trust may sustain a honesty equilibrium, again following Dasgupta, I construct a special kind of game known as the Assurance Game, henceforth denoted AG |Sen, 1967; 1973; 1985; Runge, 1981; 1984; 1986; Dasgupta, 1988; also Ullmann-Margalit, 1977: 41; Collard, 1978: 12-13,36-44, 80-89; Field, 1984: 699-700; Taylor, 1987: 18-20, 38-39,139-40; Levi, 1988: Ch. 3^. The situation it portrays is radically different from the PD inasmuch as the actors now have a predisposition towards co-operation (honesty) in the sense that they derive a positive utility from it: they therefore obtain more satisfaction from cooperating with others than by free riding on their efforts, perhaps because they have to reckon with the sense of grievance that the others will almost certainly feel if they behave in an opportunistic way |Sugden, 1984: 775^. As a result, there is no more a dominant (free rider) strategy because the particular outcome 'depends crucially both on prior expectations and on a preference for coordinating one's own actions with the actions of others' |Runge, 1984: 158^. To the extent that the structure of mutual expectations plays a crucial role, multiple equilibria become possible and the Pareto-optimal outcome is one of them. The AG is thus particularly interesting because it provides an escape from the counterintuitive result that free riding is the unique possible equilibrium in a single-period framework.

To make things clearer, let us consider the following symmetric, two-person one-shot game.(4)

FIGURE 1                                                             
                                                                     
THE ASSURANCE GAME                                                   
                                                                     
Player 2                                                             
                                                                     
                    H        C                                       
                                                                     
            H     20,20     5,15                                     
                                                                     
Player I                                                             
                                                                     
            C     15,5     10,10                                     

I assume that in the large group forming a market society individuals meet pairwise at random in every period to transact and they cannot tell if they have met before. Anonymity therefore prevails and reputation cannot be lost (a person's past record of choices is not known by anyone). Each actor can choose between two strategies: to be honest (H) or to indulge in some cheating (C), and choice strategies are determined simultaneously. As is evident from the above payoff structure, each player would prefer to transact honestly if the other were to choose to be honest, but he would prefer to cheat if the other were to act dishonestly. In other words, when dealing with honest people, it is nice to be honest, but it is unpleasant to be exploited by dishonest transactors. Moreover, each one prefers both to choose H rather than C: bilateral honesty is deemed preferable to bilateral cheating.

It is immediately apparent that, for each party, the best policy depends on what he thinks the other will do. In fact, optimal choice of one player is H if the probability that the other player will choose the same strategy is assessed by him to be in excess of 1/2, and his optimal choice is C if this probability is less than 1/2.(5) Thus, while there is no certainty that the game will equilibrate at the more favourable of the two (Nash) equilibrium points, HH and CC, this can happen if the transactors have enough confidence in each other, the degree of trust being measured by the probability that the other chooses H instead of C. It may be noted that the degree of trust needed to yield co-operation is inversely related to the intensity of the agents' inclination towards honest dealing. Thus, if it is assumed that each player gets a payoff of 30 instead of 20 when both behave honestly, a player will be induced to so behave if his expectation that the other will also be honest is greater than or equal to only 1/2 (instead of 1/2).(6)

Another point deserves to be made. If all agents have an AG-payoff structure and if, when they meet pairwise, they could make sequential (instead of simultaneous) moves, honest behaviour would be generalised. This is because one of the two transactors would then take the initiative of honest acting and the other, after having observed that honest move, would be incited to follow suit. (It is precisely because the leader is fully confident that his honest behaviour will certainly be emulated by the follower that he does not hesitate to make his first move.)(7)

Let us now modify the above AG so as to allow for the possible presence in the population of a group of inveterate opportunists who are eager to exploit to their own profit any situation which they encounter: in other words, these players have a payoff structure characteristic of the Prisoner's Dilemma (PD). Let us consider the following modified assurance game where type 1 players are given to honest urges while type 2 players are opportunists:

FIGURE 2                                                             
                                                                     
A MODIFIED ASSURANCE GAME WHERE OPPORTUNISTS ARE PRESENT             
                                                                     
          Type 1 player          Type 2 player             Type 2 pla
yer                                                                  
                                                                     
             H       C                 H       C                 H   
    C                                                                
                                                                     
Type I   H 20,20    5,15   Type 1  H 20,12    5,15   Type 2  H 12,12 
   5,15                                                              
player   C 15,5    10,10   player  C 15,5    10,10   player  C 15,5  
  10,10                                                              

The payoff structure of the game is essentially the same as in Figure 1 except for the fact that when type 2 players deal honestly in their exchanges with type 1 players, they get only 12 (all that matters is in fact that this payoff be strictly smaller than 15). There are three kinds of possible encounters in this modified AG game. Note that the first type of encounter (would-be honest agents meet together) is strictly equivalent to the game depicted in Figure 1. Further, it is worth emphasising that the game embodies an assumption of imperfect information. Indeed, no player knows ex ante whether he is dealing with an honest or a dishonest transactor; only ex post can he know the answer to that question. It is immediately apparent that type 2 players (the opportunists) have a dominant strategy, which is to cheat. As for type 1 players (who are a priori inclined towards honesty), their strategy will again depend on the state of their expectations regarding others' inclination to make honest dealings. Thus, assuming that P stands for the proportion of type 1 players in the population -- and (1-P) for the proportion of type 2 players -- type 1 players will choose to behave honestly provided that they expect P to be at least equal to 1/2. If P is assessed to be less than 1/2, they will opt for cheating which will be universal practice in the society.

The main conclusion that emerges from the above modified AG is the following: for honest dealings to prevail on a large scale in an anonymous society, it is not sufficient that a significant majority of people deem honesty preferable to dishonesty but it must also be the case that these people feel confident that their bent for honesty is shared by many others too. Therefore, if there are many would-be honest agents who however believe that many people around them are not similarly given to honest urges, no honesty equilibrium will arise. In addition, an important result obtained under the simple AG continues to hold: the degree of mutual trust required to cause an actual display of honesty among the agents predisposed towards it is inversely related to the intensity of this predisposition. Finally, it bears emphasis that in the Modified AG the emergence of honest behaviour is not necessarily precluded even though P is (significantly) smaller than one. In other words, this model can account for the widely observed fact that there is a certain amount of fraud and economic crime in all successful market economies: regular crookedness, provided that it is kept within tolerable bounds, does not prevent market transactions from taking place.

Dynamic considerations: Now, the question is not only how, or more precisely under what conditions, honesty in economic transactions can be established, but also whether it can be sustained over time. The easiest case obtains when everyone in the society has a bent for honesty (P = 1) and this fact is common knowledge. We then have a stable honesty equilibrium (of course, if P = 0 and this is commonly known, we have a stable cheating equilibrium: everyone cheats and continues to cheat). The issue becomes much more complex as soon as one assumes that P lies between zero and one. As pointed out by Dasgupta, to know which stationary (or steady-state) equilibrium is going to prevail in the long run (if at all the society gravitates towards a stationary equilibrium), we need more information about the dynamics of expectation formation, which clearly requires the support of detailed historical knowledge |Dasgupta, 1988: 59^. In a dynamic setting, indeed, the decision whether or not to act honestly requires continual re-evaluation of the probability that others will also behave honestly based on concrete experiences in past rounds: not only do expectations affect honest behaviour, but over time honest behaviour affects expectations |Runge, 1984: 170-71^. To illustrate the importance of the dynamic aspects of the (modified) assurance problem, let us consider the rather straightforward following scenario. Initially, P is high enough to incite type 1 players to give co-operation a try. The experiences of all agents are common knowledge so that expectations at each period adjust in the direction of the past period's average behaviour. Under these conditions, honest equilibrium behaviour is a stable expectation: a slight departure from honest equilibrium expectation will bring the expectation back to equilibrium |Dasgupta, 1988: 58^. This is clearly an ideal situation where each agent knows what to expect of others in the relevant future.

Let us now continue to assume that type 1 players start by behaving honestly. However, people's experiences are no more common knowledge -- information is imperfect because players are only aware of their own personal experiences (and perhaps also of that of neighbours and relatives) -- and people's expectations are vulnerable to bad experiences that prove honesty to be too costly a strategy. In other words, if type 1 players have an uninterrupted run of bad experiences (say, during x consecutive transactions), they turn cheaters.(8) This will actually be the case for a number of them. Other type 1 players will have mostly good experiences and will therefore remain co-operators. On the other hand, defectors remain defectors. In these circumstances, the proportion of defectors in the population rises, thereby increasing the probability that cooperators meet defectors and decide to free ride following the disappointment of their expectations. Trust may thus gradually unravel, leaving ultimately no co-operator in the society. In the above example, the kind of trust exhibited by (would-be) honest agents reflects an attitude of initial openness to evidence, these agents acting as if they trusted the others, at least until more stable beliefs can be established on the basis of further information |Gambetta, 1988b: 234^. In other words, trust corresponds to an attitude of initial predisposition to honesty subject to revision over time depending upon the degree of fulfilment of expectations. In this case, there is evidently no a priori assurance that honesty is sustainable in the long term even if a majority of agents are honest to start with.

Fortunately, this scenario is not the only one that can be imagined. Alternative processes of expectation formation can possibly avoid the gradual unravelling of trust. This could happen, for instance, if cheaters feel guilty to continuously cheat amidst a large number of honest dealers, or if would-be honest agents who have turned cheaters after a run of bad experiences are ready to start again behaving honestly after a run of good experiences (in the latter case, we need to know the precise mechanism of expectation formation to determine whether honest behaviour is a stable equilibrium). On the other hand, additional possibilities for sustaining honesty arise when the payoff structure of the game is allowed to be modified. For example, the unravelling of trust and the erosion of honest practices can be avoided if honest agents feel vengeful enough and are able to punish defectors at low cost to themselves. Protection of trust is still better guaranteed if honest agents punish not only defectors who have harmed them personally but also those who have been found to harm other people with whom they somehow identify.

To refer to another approach to the above question, Guth and Kliemt |1994^ have recently shown with the help of evolutionary game model-ling that honest people (or people 'with a conscience') need not be completely wiped out of a population invaded by non-trustworthy mutants. To put this result in perspective, it has to be borne in mind that in evolutionary models of biological competition, patterns of behaviour that have proven relatively successful in the past are more likely to survive and reproduce. In the human world, the evolutionary approach implies that a deviant play, if successful (that is, if it brings a relatively high payoff), tends to be repeated and imitated |Sugden, 1989: 92^.(9) Notice carefully that these patterns do not correspond to genuine strategies in the sense of game theory since they result from myopic choices of agents who do not clearly anticipate the long-term consequences of their present decisions including all the possible reactions of other players.

So much for the general framework used by Guth and Kliemt. Turning now to the more specific assumptions underlying their evolutionary games, the following have to be noted. First, there are two types of players: trustworthy or fair players (with an AG-payoff structure) and opportunistic or unfair players (with a PD-payoff structure), that is, exactly the same types of players as those assumed to play in Figure 2. Second, these two types of players indefinitely interact, yet they are matched randomly at each round of the game and they play independently of their memory of past or their expectations of future interactions with either the same or other players. Third, the games in which they play are 'games of trust', meaning that social interaction is characterised by a sequence of moves in which one player moves first and another second. Four, each player, regardless of his type, will be in the position of the second mover with probability 1/2.

What the authors essentially show can be summarised as follows. In the first place, if each player is perfectly informed about the other player's type, the fair type of player fare better than the unfair type and, consequently, starting with any population that contains both types the fair type will eventually eliminate the unfair type. In more technical terms, there is exactly one evolutionarily stable strategy in a game of trust with complete information, namely 'playing fair'. (A strategy is evolutionarily stable when it can resist to a mutant invasion by another strategy in the sense that it can prevent the latter from establishing itself permanently in the population.)(10) This result is actually not surprising given the fact that, if an unfair player has the first move in an encounter with a fair player, he has an incentive to play fairly rather than unfairly. The assumption of a sequential game is therefore crucial in this context.

Guth and Kliemt then relax the assumption of perfect information: indeed, such an assumption does not square well with that of random matching which better fits with a process of anonymous interaction of a large number of players. They instead assume that type information is lacking. Under these conditions, the result obtained is less encouraging. As a matter of fact, the fair type can no more resist an invasion of unfair 'mutants' because the latter would have a higher reproductive success once they enter a population composed exclusively of fair types. Yet, once the proportion of fair individuals shrinks to some value lower than a certain threshold (corresponding to the ratio of the payoff accruing to each player when they both behave dishonestly to that accruing to them when they both behave honestly), there would be no evolutionary pressure in favour of unfair types anymore. Things appear much worse if the assumption is made that players can make slight mistakes or non-rational choices once in a while. In this case, indeed, only populations with no fair individuals shall be evolutionarily stable.(11) Hence the (provisional) conclusion that 'the disposition to develop a conscience that is sufficiently strong to motivate fair behaviour shall not succeed in evolution' |Guth and Kliemt, 1994.' 18^.(12)

Fortunately, the last assumptions do not necessarily reflect correctly what obtains in the real world. In a last step, Guth and Kliemt therefore assume that a (perfectly reliable) technology is available at a cost to identify the other player's type. Note that, in games of trust, it is evidently the information about the trustworthiness of second movers that is crucial. The results are the following. A monomorphic population with only fair players can never be evolutionarily stable while a monomorphic population comprised of opportunists only will always be evolutionarily stable.(13) Yet, if the cost of the detection technology is not too high, there are two evolutionarily stable populations that are characterised either by the complete absence of fair players or by a positive proportion of such players. More precisely, whenever the proportion of fair players falls below a certain threshold P*, fair individuals will be driven out of the population. On the contrary, once fair types have somehow reached that critical proportion, trustful co-operation can survive: as a matter of fact, their share in the population will then increase up to another threshold P** which is less than one (unfair types cannot be eliminated). The values of these two thresholds are as follows:

P* = 2c/r - s and P** = s - 2c/s

where c is the cost of the detection technology, r the payoff accruing to both types of players if both behave honestly and s the payoff accruing to them if both behave dishonestly. We also have 0 |is less than^ s |is less than^ r |is less than^ 1, bearing in mind that, when one player 'exploits' the other, he gets 1 while the latter gets only zero. As is evident from the above definition of P*, the requirement necessary for the evolutionarily stable survival of fair types becomes more severe as r gets nearer to s, that is, when the advantage of mutual honesty over mutual dishonesty is less significant. Also note that the evolutionarily stable composition of the mixed population (containing both fair and unfair types), P**, varies inversely with the cost of the detection technology. Finally, when c = 0, P* = 0 and P** = 1, implying that a monomorphic population with only fair type players is evolutionarily stable as stated under the first result above. (To say that identifying the type of the other player is a costless process is tantamount to assuming perfect type information).(14)

An important implication of the last set of results is, of course, that the starting point of the dynamic process is essential. If the detection technology is not too expensive, -- if it is less than s/2(1 - s/r) -, it depends on the starting point P |degrees^ whether the population composition will converge to P = 0 or to P = P**. Both is possible. 'If it so happens that we have a "good start" there will be a more or less happy ending if not so not' |Guth and Kliemt, 1994: 35^.

Notice carefully that the assumption of sequential moves is absolutely crucial. If it is replaced by that of simultaneous moves, much more pessimistic conclusions obtain. Consider an evolutionary model with the following characteristics: there is random matching; moves are simultaneous; a reproduction rule ensures that players with higher payoffs multiply (whenever a player gets a cumulated payoff of x, he gives birth to another player possessing the same features as himself); expectations about others' predisposition to honesty are adjusted upwards after a good experience and downwards after a bad experience (according to some predetermined adjustment rule). It can then be shown, using a simulation procedure, that, even if such expectations are initially very high and the initial proportion of fair players is also very high in the population, the incidence of co-operation dwindles and reach zero after many rounds (partly because unfair players multiply rapidly and partly because fair players turn unfair). In actual fact, it is only under assumptions extremely favourable to 'fair play' (honesty) -- the initial proportion of fair players is extremely large, and the sensitivity to good experiences is extremely high while that to bad experiences is extremely low (expectations about others' predisposition to behave honestly are quick in adjusting upwards but slow in adjusting downwards) -- that honesty may survive over the long run without being nevertheless able to eliminate dishonest players altogether |Gaspart, 1994^.

Total pessimism is unwarranted, however. It is easy to show that, even with simultaneous moves, results are much more favourable to cooperation (honesty) when it is alternatively assumed that players can somehow, even at a positive cost, identify the type of the other players. Under these circumstances, fair players can indeed match with other fair players, leaving the unfair players to deal with one another |Frank, 1988: 56-63^.

The Role of Generalised Morality

The nature of moral norms: Leaving aside the characteristics of the players' moves (simultaneous or sequential), there are a number of interesting conclusions which emerge from the above discussion. Honesty is especially likely to be established and sustained on a significant scale if (1 |degree^) a large number of people have a preference for honesty to start with; (2 |degrees^) they also have sufficient trust in others' predisposition towards honesty; (3 |degrees^) their bent for honesty is strong enough not to be easily discouraged by bad experiences while it is easily reinforced by good experiences; (4 |degrees^) cheaters are subject to strong guilt feelings when they are free riders among many honest dealers; and/or (5 |degrees^) honest people are willing to sanction breaches of honesty conventions even when their own interests have not been harmed by the observed breach (and despite the fact that private sanctioning activities will probably not bring them any direct reward in the future). Note that conditions (4) and (5) imply that the structure of the modified AG depicted in Figure 2 is being altered. With respect to condition (2), it is worth bearing in mind that the stronger the preference for honesty the smaller the degree of trust needed to yield the honesty outcome or, in terms of the evolutionary games of trust, the greater the likelihood that honest individuals may form an evolutionarily stable proportion of the population (when r is large relative to s, P* is small). What I shall argue now is that the fulfillment of the five conditions above largely depends on the prevalence of moral (other-regarding) norms in the society.(15)

Norms are expectations about one's own action and/or that of others which express what action is right or what action is wrong |Coleman, 1987: 135^. The concept suggests a standard of conduct which people believe they ought to follow lest they should expose themselves to some way of sanctioning or unpleasant experience. Obedience of the norm will occur when the sanctions or discomfort are sufficiently great and sufficiently certain to make disobedience less immediately attractive than obedience |ibid.: 141-2; 1990: 242-3^. For patterns of behaviour to be sustained by norms, a society (and the underlying social consensus) must therefore exist to impose sanctions on norm violators. This can be done through a central agency acting as an external norm-enforcer, through agents inflicting sanctions upon one another in a decentralised manner, or via a self-policing mechanism. We know from previous discussion that the first two solutions are questionable: the first one is likely to entail considerable costs while the second one is more appropriate for small groups than for large societies where reputation effects cannot be sufficient to police behaviour and where private sanctioning activities do not bring any future reward to the punishing agent. We are therefore left with the third solution which implies that external monitoring and sanctioning devices (whether formal or informal) can be actually dispensed with. Moral norms, understood as rules that are at least partly internalised by the agents (in Freudian terms, they form their superego) and that prompt them to take others' interests into account, provide such a self-policing mechanism.(16)

Internalisation of standards -- whereby the actor exercises first-party control |Elickson, 1991: 126^ -- may be said to arise when an individual actually conforms because of a personal attitude about the act itself, that is, when conformity becomes a motive of its own because it is intrinsically rewarding or because deviation is intrinsically costly |Weber, 1971: 22-3; Opp, 1979: 777, 792; 1982: 146; Jones, 1984: 89; Taylor, 1987: 13^. Hence 'internalization refers to the aspect of the process of socialization through which attitudes, values, and behaviour patterns come to be maintained even in the absence of external rewards or punishments' |Jones, 1984: 89-90; also Aronfreed, 1968; 1969; 1970: 104; Bergsten, 1985: 115; Coleman, 1990: 245, 293^. As a result, moral norms are followed even when violation would be undetected, and therefore unsanctioned, because the moral act -- which appears to be in conflict with the immediate or direct interests of the actor himself -- is valued for its own sake |Griffith and Goldfarb, 1988: 22; Elster, 1989a: 131, 1989b: 104^.(17)

Now, an important lesson from developmental psychology is that moral behaviour and the ability to empathise emerge hand-in-hand with the maturation of specific emotional competencies.(18) In this maturation process, identification obviously plays a crucial role since a failure to follow the standards set (consciously or not) by reference persons is bound to generate the painful feeling that one is unable to meet their expectations and, thereby, to deserve their love or respect. Thus, the principal claim of Kagan |1984^ is that moral norms are actually supported by a limited number of simple, highly uniform emotional capacities. According to him, the main motivating force behind moral behaviour is the desire to avoid feelings of guilt and shame which are themselves the combined outcome of unpleasant emotions (anxiety, empathy, responsibility, fatigue/ennui, uncertainty).(19) 'Bad' or unpleasant feelings stirred by violations of the prevailing moral norms are the ingredients of a 'tortured conscience' which tend to deter many people from breaking these norms. Such deterrence would not occur if violating norms was just felt as a mistake or a lapse from rationality |Elster, 1989a: 188^. Moreover, as found in numerous experiments, subjects induced to commit some transgression (or to believe that they have transgressed) are more likely than nontransgressors to engage in altruistic or compensatory behaviour. This is apparently because they are eager to repair their self-conception or self-presentation so as to convince themselves as well as others of their moral worthiness |Shott, 1979: 1327; Darley and Latane, 1970: 100-101^.

Notice that, in the above scheme of analysis, emotions or passionshave a positive role to play in society.(*) This is at variance with theposition adopted by Adam Smith in The Theory of Moral Sentiments:there, we are told that passions (fear and anger on the one hand, and 'the love of ease, of pleasure, of applause, and of many other selfish gratifications' on the other hand) are so many drives that are apt to mislead man into mischievous actions while their control enables him 'upon all occasions to act according to the dictates of prudence, of justice, and of proper benevolence' |Smith, 1759: Part VI, Sect. III, 238^. What Smith appears to believe is that, when man reasons coldly and practices self-command, he can think of the long term or of the social consequences of his behaviour (at least if he is a 'wise' man) while, on the contrary, when he is given to the urgent drives of passions, he is unable to see beyond his immediate short-term interests and he may thus be sometimes seduced 'to violate all the rules which he himself, in all his sober and cool hours, approves of' |ibid.: 237^. What I argue instead, following a line suggested by Frank |1988^, is just the opposite: emotions are susceptible of leading people, almost unconsciously, to overcome the temptation to give in to short-term considerations and to therefore act in accordance with their long-term interests.

This being said, it would be wrong to believe that, once they have been properly internalised, moral norms are completely compelling with the result that decision outcomes are mechanically determined.(20) As has been aptly noticed by Elster, moral beings are usually outcome-insensitive with respect to benefits but not with respect to costs: the costs of co-operation (honesty) may be so high as to offset the call of duty. Furthermore, they are somewhat sensitive to benefits in the following sense: if they do not consider the likely impact of their own co-operation (honesty), they pay attention to the impact of universal co-operation (honesty). It is plausible that the strength of their feelings of duty depends on the difference between universal co-operation (honesty) and universal noncooperation (dishonesty): 'The smaller the difference, the lower the voice of conscience and the more likely it is to be offset by considerations of cost' |Elster, 1989a: 193^.

Generalised morality and the market order: From the above discussion, it is evident that inculcation of moral norms involves much more than purely cognitive learning. This is particularly true of primary socialisation which an individual undergoes in childhood since it takes place under circumstances that are highly charged emotionally |Berger and Luckmann, 1967: 149-57^. Note carefully that primary socialisation creates in the child's consciousness a progressive abstraction from the roles and attitudes of concrete significant others (usually the parents) to roles and attitudes in general, implying that the child becomes able to identify 'with a generality of others, that is, with a society' |ibid.: 152-3; Lane, 1991: 82^.(21) This is an important aspect in so far as any moral rule includes an element of conceptual generality that involves the capacity to recognise the claims of others and to impose such rules both on oneself and on others similarly situated |Griffith and Goldfarb, 1988: 22-3^. Moreover, in so far as it is rooted in the idea that there exists a community of people linked by solidarity ties, a strong moral attitude may be associated with the belief that most others are also behaving morally, although the attitude itself is not formulated in conditional terms (see Bromley and Chapagain |1984^, for such an interpretation of moral behaviour in a Nepalese village). The underlying principle of 'unconditional commitment' (Sugden) -- each individual behaves the way he would wish others to behave, regardless of which way they actually behave -- may appear too strong to form the basis of practical morality, however |Sugden, 1984: 774-5^. A less stringent moral principle is that of reciprocity: 'you behave the way which you would like the others to behave, but only if they actually meet this expectation' |Sugden, 1984: 775; Elster, 1989a: 214^. A striking illustration of this principle at work in an Indian village is provided by Bagchi |1992^.

Now, for generalised morality to be a 'social capital' (Coleman) capable of sustaining order in the marketplace, it is essential that concern for others or ability to see things from another's viewpoint be based on identity or loyalty feelings towards a large reference group actually encompassing all the relevant market transactions. This is precisely the condition which Granovetter does not think it possible to satisfy when he contends that norms of limited-group rather than generalised morality can tame fraud and deceit: in his own words, networks of relations, rather than institutional arrangements or generalised morality 'are the structure that fulfils the function of sustaining order' |Granovetter, 1985: 491^. None the less, the ability to recognise the claim of a large generality of others is clearly present in the Christian ethical principle according to which we ought not to do to other people what we would not like them to do to us; or in the Kantian generalisation principle according to which one ought to abstain from any action that would threaten to disrupt social order were everybody to undertake it or that one would not be prepared to see everyone else adopt (an action is morally possible only if it can be universalised without self-contradiction).

The ability to put others on a similar footing with oneself is at the heart of classical republicanism, a political doctrine (initially formulated by Nicolo Machiavelli and several of his contemporaries and later taken over by Alexis de Tocqueviile) which is centered around the notions of a well-ordered republic and public-spirited commitment to the common good. The emphasis on virtues of civic humanism and on the need to withstand the dictates of private interest whenever they undermine the public good were expected to enable men from different kin groups to trust each other, thus making possible the operation of a 'commercial society' |Pagden, 1988: 139; McNally, 1988: 40-43^. The communal republics of Northern Italy during late medieval times served as a reference case for many founders of this political doctrine. Thanks to a first and fervent allegiance to their own city-state, members of these mini-republics are reported to have shown an unparalleled degree of civic engagement, as a result of which trust could be extended beyond the limits of kinship further than anywhere else in Europe in this era |Putnam, 1993: 127-31^.(22) On the other hand, according to Skinner the social ethos of Calvinism can be actually seen to constitute a special case of classical republicanism |Skinner, 1974^. If this view is accepted, it might then be maintained that 'it was not a specifically Calvinist or Puritan work ethic which encouraged economic take-off in the Protestant cities of the seventeenth century, or Christianity itself in the medieval cities of north and central Italy, but the secular ethic of classical republicanism' |Pagden, 1988: 139^. Such a perspective invites us to pay attention to Weber's little known thesis that 'the great achievement of the ethical religions' had been 'to shatter the fetters of the kin', an outcome which could not materialise in China where 'the fetters of the kinship group were never shattered' owing to the absence of a similar ideology (Weber |1951: 237^ quoted from Pagden |1988: 139^).

There is yet another fine observation by Weber that is worth mentioning here because it also allows us to link up the above conclusion about the achievement of generalised morality in western Europe in modern times with what I have said earlier (in Part I) about the limited domain of trust in traditional societies. Weber's point is indeed that the universal diffusion of unscrupulousness in the pursuit of self-interest was far more common in precapitalist societies than in their more competitive capitalist counterparts (Weber |1970^ quoted from Gambetta |1988b: 215-16^). While normative behaviour based on the principle of reciprocity (I cannot expect others to deal honestly with me unless I am honest with them) was restricted to real kith and kin in the former societies, security of expectation or assurance has been provided by the pervasive influence of a code of generalised morality in the latter. What deserves to be emphasised is that, when such a code is provided or supported by an universal religion, the generality of others to which it applies is vast. This avoids the major shortcoming of state-driven consensus ideologies which are often based on patriotic identification and on projection of aggression against outsiders, thereby resulting in a restriction of the market size to national boundaries. Societies led by these ideologies tend to be 'not merely bureaucratized and heavily internally-policed but also paranoid in their foreign relations and militaristic to a high degree' |Badcock, 1986: 182^.

If the foregoing analysis is correct, the crucial factor behind 'the rise of the western world' in the modern era is not the 'cult of individuality', or not that cult alone, but a much more subtle combination of factors: in western Europe since the middle ages the emancipation of the individual (within the framework of national spaces) from erstwhile networks of social and political allegiance went hand in hand with the development of generalised morality in which abstract principles or rules of conduct are considered equally applicable to a vast range of social relations beyond the narrow circle of personal acquaintances.(23) It is no doubt true that the borderline between limited-group and generalised morality is somewhat blurred because the latter is never absolute but always restricted too. It is thus obvious that western societies have typically ceased to apply their moral-ethical principles once they came into contact with non-European peoples on the periphery of 'their' world. The history of slavery and of colonial exploitation is there to remind us of this basic fact: for example, it was considered morally legitimate by Europeans to cheat American Indians on the grounds that they were lazy, irrational, and unwilling to repay their debts |Ortiz, 1967: 203^. This being said, if limited-group morality is understood as morality restricted to concrete people with whom one has close identification while generalised morality is morals applicable to abstract people (to whom one is not necessarily tied through personal, family, or ethnic links), there is good sense in arguing that the western world has a somewhat unique history rooted in a culture of individualism pervaded by norms of generalised morality. At least, this started to hold true for elite circles which inculcated such norms in the lower classes, usually for their own benefit. (Note indeed that moral norms are compatible with inequality as is attested by norms teaching respect for private property even when such property is unequally distributed, or norms teaching workers hard work, discipline, submission, and temperance.)(24)

The need for reinforcement mechanisms: We have seen above that primary socialisation plays a crucial role in the norm generation process. Nevertheless, moral norms are subject to erosion: they form a 'social capital' |Coleman, 1987^ and, as such, they are liable to depreciation, especially so if the cost of honesty is high or increasing. They therefore need more or less continuous reinforcement to be maintained. One such kind of reinforcement consists of what Berger and Luckmann have called secondary socialisation. Secondary socialisation is 'the acquisition of role-specific knowledge, the roles being directly or indirectly rooted in the division of labour' |Berger and Luckmann, 1967: 158^. Contrary to primary socialisation which cannot take place without an emotionally charged identification of the child with his significant others, most secondary socialisation 'can dispense with this kind of identification and proceed effectively with only the amount of mutual identification that enters into any communication between human beings' (it is necessary to love one's mother, but not one's teacher). In primary socialisation, the child does not apprehend his significant others as institutional functionaries, but as mediators of the only conceivable reality: in other words, he internalises the world of his parents not as one of many possible worlds, not as the world appertaining to a specific institutional context, but as the world tout court |ibid.: 154, 161^. This explains why the world internalised in primary socialisation is 'so much more firmly entrenched in consciousness than worlds internalized in secondary socializations'. As a matter of fact, since secondary socialisation's main function is to transmit specific knowledge (in schools, in factories and so on), the social interaction between teachers and learners can be formalised and the former are in principle interchangeable |ibid.: 154, 162^. Note that reinforcement of moral norms is particularly effective when church attendance ensures that people are continuously exposed to a moral discourse which repeatedly emphasises the same values as they were taught by their primary socialisation's agents during their childhood. The Church (both Catholic and Protestant) obviously played a central role in the process of moral norm generation and maintenance throughout modern western history. Its impact was all the more significant as (1 |degree^) it promised a considerable reward (an eternal life of absolute happiness) for all those who were ready to incur personal sacrifices by behaving in other-regarding ways; and (2 |degrees^) monitoring costs could be brought to a minimum insofar as God was thought to act as an impartial and free monitoring agent. In the words of Frank |1988: 250^:

Teaching moral values was once the nearly exclusive province of organized religion. The church was uniquely well equipped to perform this task because it had a ready answer to the question. 'Why shouldn't I cheat when no one is looking?' Indeed, for the religious person, this question does not even arise, for God is always looking.

From the above, it should nevertheless not be inferred that all (external) manifestations of religiosity and clericalism are necessarily positively correlated with cooperative (honest) behaviour or civic-ness. In certain historical or social settings, engagement in Catholic groups and adherence to Catholic rituals, for example, may be purely formal and even hypocritical with the result that no genuine commitment towards taking the others' viewpoint into account is implied. Thus, Putnam has recently argued that in Italy, at least, 'the most devout church-goers are the least civic-minded' |Putnam, 1993: 175-6; also 107, 130^.

Moral norm reinforcement must also come from the state or the rulers. As a matter of fact, morality and a high sense of public purpose among the rulers are important ingredients of a well-functioning market order not only because they ensure that rules will be properly enforced (corruption among state administrators will be held in check), but also because of the positive demonstration effect exercised by leaders with whom people have perhaps come to identify strongly. When the public attitude of these leaders exerts a significant impact on people's behav-iour, fewer laws are needed and, to the extent that they exist to regulate behaviour, people are more likely to comply. Consequently, less repressive or sanctioning resources are required to ensure their enforceability. In the next section, mention will be made of the potential role of the state as a norm activator or supplier rather than as a norm reinforcer.

Moral norms and civil society: Emphasis on the potential role of the state as a norm-reinforcing agency ought not to deflect our attention from the equally important role that a living and watchful civil society must play to support honest behaviour and to banish economic crime. Quite obviously, the contribution of civil society to norm maintenance is especially significant when the state is wanting in this respect and when its agents are perhaps themselves involved in various sorts of illegal or fraudulent practices. In terms of my scheme of analysis, a civil society helps to back trust whenever there is a sufficient number of citizens who feel vengeful enough to work towards exposing publicly the illegal acts or malpractices of both private and state agents, and towards bringing pressure to sanction them even though such actions entail significant costs to themselves (by denouncing economic crimes they produce a public good). To the extent that the state cannot be relied upon to (re)activate norms of honesty, either because it is itself corrupt or because it has no moral clout or ideological persuasive powers, the existence of an active civil society imbued with moral norms appears as an important condition of economic growth. My argument, it must be noted, does not imply that, at every stage of market development, political democracy is absolutely necessary for honest behaviour to be reasonably well maintained. It can therefore accommodate the scepticism recently expressed by Bardhan about the relationship between democracy and economic growth |Bardhan, 1993a^. Nevertheless, to the extent that the ability of the state to act as a norm supplier (through charismatic leadership and/or through organised ideological persuasion) is subject to erosion over time -- leaders pass away and ideologies tend to wear themselves out -- the argument implies that in the long run market expansion and economic growth must be paralleled by the development of civil society and the associated freedoms of political expression.

Insights from experimental (social) psychology: That moral norms are at work in modern market societies is evident not only from day-to-day experience and historical evidence but also from experiments in western (particularly US) social psychology. In a general way, experimental research (and observation of real-world situations) indicates that individuals do not exploit free riding opportunities in the manner predicted by the PD paradigm, especially because of the interaction between their perception of fairness and their own concern for fairness |e.g., Rapoport and Chammah, 1965; Eiser, 1978; Marwell and Ames, 1979; 1980; 1981; Akerlof, 1983; Bromley and Chapagain, 1984; Roth, 1988; Dawes and Thaler, 1988; Frank, 1988: Chs. 9, 11; Bagchi, 1992^. Particularly relevant to the discussion of generalised morality is the fact that reciprocal altruism (altruistic acts performed in the expectation of a future personal gain) and tit-for-tat cannot explain cooperation in many experiments simulating PD because the games are played only once or defection simply cannot be detected. In some experiments, a significant statistical correlation was actually found to exist between the amount contributed by the subjects to the provision of a public good and the level of contribution expected of others |Marwell and Ames, 1979: 1356^. When there is near unanimity of thought regarding what is fair behaviour (for instance, a fair contribution to the provision of a public good), implying that people hold common subjective expectations about the behaviour of others, free riding appears to be the exception rather than the rule |Runge, 1984: 173-5^.

With specific reference to the problem of honesty, field experiments conducted by Hornstein and his colleagues are especially illuminating.(25) They indeed revealed that, in what is typically a one-shot game, an astonishingly high 45 per cent of 'lost' wallets were returned completely intact to their owner in New York city (during the spring of 1968). Moreover, Horstein et al. were able to show that the return rate was significantly higher when the subjects of the experiments were exposed to a positive attitude of benevolence on the part of a third party. The interpretation offered by the authors is that the third party served as a role model for the subjects. A related lesson is that feelings or sentiments, not reason, motivate human decisions in situations where our own acts have a significant influence on others' well being: exposure to different kinds of persons or acts (benevolent or malevolent) evokes particular emotions which drive people to behave in certain ways |Frank, 1988: 216^. Another experiment which confirms the important function of role models (or the fact that altruism or morality is encouraged by the observation of it) is that reported by Singer |1973^. A helpless-looking woman was standing near a car with a flat tyre along the road. It was found that drivers passing this woman were more likely to come to her rescue when they previously had the opportunity to observe helping behaviour in a similar type of situation.

Such experiments would seem to suggest that role models serve as a signalling device reminding people that there are honest people around. The result would be to enhance people's trust in others' predisposition towards fair dealings. Yet, this is to neglect the emotional dimension rightly emphasised by Frank. It is actually more satisfactory to view role models as privileged agents who reactivate emotional capacities associated with primary socialisation processes. Moreover, if one still wants to cling to the rational egoist's model of the economists, one may consider that role models have the effect of increasing -- or restoring to previous levels -- the values of the payoffs attached by people to the HH outcome. Note carefully, however, that contrary to a well-established tradition in economic theorising, this latter interpretation assumes that individual preferences are not stable. It also enables us to better understand the potential role of political leaders in diffusing or reinforcing norms of honesty (see supra). Political leaders now appear as norm reactivators. When they publicly behave in honest ways, they naturally arouse in people the emotions associated with that type of behaviour provided, of course, that people have sufficiently strong feelings of identification with them.(26) Thus, for example, during the years after the revolution, Chinese leaders appearing in Mao suits and living according to sober fashions both at work and at home motivated people to earn honest incomes. On the contrary, the rise of corruption practices among the elite, particularly since the liberalisation reforms |Riskin, 1987: 336-7^, has probably contributed to lower the level of morals in Chinese society. Also, many African leaders who display considerable amounts of wealth fraudulently acquired tend to drive people to cheat in private dealings, to evade taxes and to embezzle public funds.

A process of norm activation is apparently also at work in small groups where communication (and exchange of promises) takes place, even if only for a short period, between subjects who have met at random for the first time. In such conditions, indeed, as various experiments have shown, communication tends to raise the cooperation rate significantly, especially so if intra-group discussion leads to universal promise-making |e.g, Dawes, McTavish, and Shaklee, 1977; Dawes and Thaler, 1988; Ostrom, Gardner and Walker, 1994; Chs. 7-9^. It would nevertheless be a hasty step to infer from the above results that people are given only to limited-group morality. In point of fact, there is a fundamental difference between deliberately limiting co-operative behaviour to a restricted circle of well-identified persons with whom one has ex ante empathy feelings on the one hand, and being a priori opened to co-operation with a large number of people provided that some sort of acquainting process occurs during which the 'abstract other' becomes somewhat more familiar, on the other hand. Communication, I would suggest, performs at least three functions. First, it enhances trust in others' predisposition to co-operate with us, especially so when communication is accompanied by a verbal exchange of promises (many people are particularly prone to believe others when the latter say pleasant things or point to attractive prospects).(27) Second, communication triggers our ability to adopt the others' viewpoint. And, third, it intensifies guilt feelings since the potential victim of our free riding has been transformed from a mere stranger into a real person. If this interpretation is correct, one important lesson from the above-reported experiments about the role of communication is that they tend to confirm a previous observation according to which moral norms are not completely compelling. Rather than a habit leading to mechanically-determined outcomes, generalised morality appears as a predisposition towards co-operation that may be activated by certain favourable circumstances and remain latent when these circumstances are absent.

Conclusion and Final Considerations

It is now the time to sum up the main argument. At the beginning of the previous subsection a series of five conditions have been stated that tend to make honest behaviour more likely to be established and sustained (especially so if they are all simultaneously satisfied). What has been shown is that the prevalence of moral norms in a society favours the emergence of these five conditions. Thus, when such norms (which are by definition other-regarding and largely internalised) are well established and effectively sustained (through appropriate secondary socialisation processes), people tend (1 |degree^) to adopt the others' viewpoint when making decisions that may harm others' interests and to feel internally rewarded when behaving in other-regarding ways; (2 |degrees^) to be confident that others will abide by the same code of good behaviour as themselves; (3 |degrees^) to cling to this code even when they had unpleasant experiences in which they were 'suckers'; (4 |degrees^) to feel guilty after they have (perhaps mistakenly) deviated from the moral rule; (5 |degrees^) to feel vengeful and willing to punish detectable free riders even when their own interests have not been directly harmed or threatened. Note that, through the first effect, moral norms transform the PD game where universal dishonesty is the only (Nash) equilibrium into an AG where honesty becomes a possible equilibrium. Through the second effect, they create trust and thereby ensure that the 'good', Pareto-superior equilibrium is selected. As for the last three effects, they help establish the dynamic sustainability of this equilibrium.

The fact that moral norms are typically inculcated in early childhood when they are strongly associated with the maturation of specific emotional competencies is important with respect to almost all the aforementioned points. In particular, this characteristic of moral norms helps resolve the second-order problem of monitoring and punishment (what incentives individuals have to monitor dealings and impose sanctions on defectors since such activities are themselves public goods?). In the tradition of Smith's Theory of the Moral Sentiments |1759^, rational assessment indeed appears as merely one of many inputs into the psychological reward mechanism and 'rational calculations often lose out to other, more basic forms of reinforcement' |Frank, 1988: 197^. Thus, vengefulness is an emotion that may easily drive persons to punish defectors even at a significant positive (short-term) cost to themselves, particularly so if they are deeply (morally) shocked. In this perspective emotions are commitment devices which have the effect of breaking the tight link between utility-yielding goals and the choice of action |Sen, 1985; Frank, 1988^.

It is also useful to recall that the monitoring problem does not even arise if moral rules are backed by religious beliefs according to which God knows everything about all our actions (and thoughts). Under these circumstances, honesty can be sustained even in the presence of high costs of fraud detection since a free monitoring device is actually available. It may be further noted that, insofar as religious beliefs imply the hope that eternal life (or a better future life) will be accorded to all righteous people, the latter's expectations will be robust. In other words, believers continue to adhere to the moral code even though they have possibly been 'suckers' on repeated occasions (see the third point above). By continuing to behave morally despite unpleasant experiences, they may even be convinced that they deserve special attention from God.(28) When such religious beliefs do not exist or are not shared by a significant majority of people, it is all the more important that moral norms are being regularly reinforced by other agents of secondary socialisation or by role models (such as political leaders) lest they should gradually erode leading to a vicious circle of unravelling trust.

III.DYNAMIC CONSIDERATIONS ABOUT THE EMERGENCE AND EROSION OF MORAL NORMS

For largely impersonal exchanges to take place on a significant scale, a social consensus must exist in the society that ensures mutuality of agreement on fair behaviour in economic dealings as well as on the structure of basic (property) rights which, as pointed out by Buchanan, actually define the entities who enter negotiations |Buchanan, 1975: 18^. Such a social consensus or mutuality of agreement is best based on moral norms of a generalised kind. These norms fulfil the function of imposing "impartial' constraints on the pursuit of individual interests, constraints which are socially desirable in serving interests that individuals share as members of a social community' |Vanberg, 1988: 3^. Now, to have a theory of norms, we should know how they arise, how they are maintained, how they change over time (how they vanish and how they are displaced by other norms), and whether and how they can be manipulated, all questions which are essentially unanswered to this date. Clearly, these are questions of considerable complexity and around which there are likely to be enduring debate and heated controversies for a long time to come. Depending on the type of answer provided, different approaches will be suggested to central problems such as the one raised in this article.

Emergence of Moral Norms

Insights from eighteenth-century philosophy: The enormous stake involved in the choice of approach to the dynamics of norms can be illustrated by considering the issue of norm emergence. Two radically opposite views are possible: they form the elements of an important controversy which has its roots in eighteenth-century philosophy. The first of these views is profoundly optimistic and is grounded in the well-known evolutionary approach to institutional change. In a pioneer attempt made several centuries ago, David Hume proposed a remarkably articulate formulation of the evolutionary theory when he tried to explain the process of emergence of social order in a market economy.

Hume actually believed that public good cannot be established unless individuals are driven not only by selfish passions but also by a 'moral sense' (a view inherited from Hutcheson). Yet moral behaviour depends upon rational considerations and, in an age of cultural and scientific progress, individuals cannot fail to see the need for private property, law, and government. Far from contributing to corruption and degeneracy, the development of 'commercial society' can be expected to pave the way for morality, justice, and good government insofar as it goes hand in hand with moral and political progress |McNally, 1988: 167-8^. At this point, Hume's ideas as contained in his Treatise of Human Nature |1740^ deserve to be detailed more fully. This is done below by quoting at some length the paraphrased description recently proposed by McNally |1988: 168-9^:

In the Treatise Hume accepts that self-love is the origin of law and government. Nevertheless, since 'the self-love of one person is naturally contrary to that of another', competing and conflicting self-interested passions must 'adjust themselves after such a manner as to concur in some system of conduct and behaviour'. After individuals discover that unbridled selfishness incapacitates them for society, 'they are naturally induc'd to lay themselves under the restraint of such rules, as may render their commerce more safe and commodious'. As rules of social regulation are developed, they become customary and are passed on to future generations. Eventually people come to cherish the rules which hold society together. They develop a sense of sympathy for those who observe social norms. Moreover, they come to model their behaviour in such a way as to be worthy of the sympathy and approval of others. Through custom and education, then, individuals develop a love of praise and a fear of blame. For Hume, moral principles are not innate or providentially inspired. They are practical rules developed in the course of living in society; morality refers to the norms and conventions which prevail there. These norms and conventions can be said to enter into the commonsense view of the world most individuals acquire . . . |Sympathy^ is a capacity derived from experience and modified as the customary rules of social life change. Sympathy has a rational dimension; it derives from the individual's understanding of the necessity for norms of conduct and behaviour. Thus, although 'self-interest is the original motive to the establishment of justice', as society develops it becomes the case that 'a sympathy with public interest is the source of the moral approbation, which attends that virtue'.

For Hume, property, law, and government are therefore the outcome of the evolution of human society.(29) Through the experience derived from their mutual interactions in the (nascent) market economy -- and not through the mental representation of abstract principles of duty or the highest good |Da Fonseca, 1991: 87-90^ -- individuals come to see the necessity of, and to accept the constraints imposed by, those institutions and conventions which preserve the social order. Men 'cannot change their natures. All they can do is to change their situation ...' and 'lay themselves under the necessity of observing the laws of justice and equity, notwithstanding their violent propension to prefer contiguous |short-term^ to remote |gains^' |Hume, 1740: Book III, Part II, Sect. VII, 537^. 'Public utility' becomes the basis of moral decision and, as Hume put it in his Enquiry Concerning the Principles of Morals, 'everything which contributes to the happiness of society recommends itself directly to our approbation and good will' (quoted from McNally |1988: 169^). The main function of rules or norms is to serve as an assurance device: 'this experience assures us still more, that the sense of interest has become common to all our fellows, and gives us a confidence of the future regularity of their conduct: And 'tis only on the expectation of this, that our moderation and abstinence are founded' |Hume, 1740: Book III, Part 11, Sect. 11, 490^. It may be further noted that Hume's analysis allows for internalised norms in so far as he holds that social rules are passed on to successive generations through education and customs (see supra).

A radically opposed and much less optimistic view than Hume's doctrine of spontaneous order of the market tends to consider that moral norms of the required kind must pre-exist in the society before a market economy can successfully develop: here, generalised morality appears as a social precondition in the strict sense of the term. Thus, for example, Edmund Burke held the opinion that 'the expansion of commerce depended itself on the prior existence of "manners" and "civilization" and on what he called "natural protecting principles" grounded in the "spirit of a gentleman" and "the spirit of religion"' (Hirschman |1987: 160^ referring to Burke |1790: 115^). If this view is followed, therefore, the market economy cannot be expected to gradually and unconsciously generate the social conditions upon which its viable existence rests. These conditions must be present in the society before the market economy can arise.

Adam Smith had a more ambivalent attitude even though, on the whole, he inclined to think in the way of his friend David Hume. That is, he essentially shared the doctrine which can be traced back to Montesquieu and which Hirschman has dubbed 'the Doux-commerce thesis' |Hirschman, 1977; 1982^. For Smith, indeed, the spread of commerce and industry enhances virtues such as industriousness, assiduity, frugality, punctuality and, more relevant to our discussion, probity. Commerce brings about moral improvement by providing honest employment and eliminating feudal institutions of dependency |Rosenberg, 1964; Hirschman, 1982: 1465; Young, 1992: 80^.(30) As is well known, in his Theory of Moral Sentiments |1759^, Smith laid much stress on the fact that all individuals have a capacity for sympathy with others, that is, they are all able, by an act of imagination, to adopt the others' viewpoint and to understand their reactions (see, in particular, Smith |1759: Part III, Ch. 1^; see also Hume |1740: Book II, Part II, Sect. VII and Part III, Sect. VI^). This capacity of sympathetic identification which enables actors to adopt the standpoint of an impartial spectator who observes situations of human interactions dispassionately is the fundamental basis of society: it supplies a system of cultural restraints in which moral checks upon the passion of self-love can be embedded |McNally, 1988: 183; Brown, 1988: 61-2, 67; Young, 1992: 73-7^. Note that Smith actually believed that the capacity to sympathise was especially noticeable among those occupying 'the inferior and middling stations of life'. This is because, being obliged to be prudent, they had to learn that the pursuit of self-interest must be held within socially acceptable bounds |McNally, 1988: 186^.

Seen from this angle, Smith shared the belief that

a society where the market assumes a central position for the satisfaction of human wants will produce not only considerable new wealth because of the division of labour and consequent technical progress, but would generate as a by-product, or external economy, a more 'polished' human type -- more honest, reliable, orderly, and disciplined, as well as more friendly and helpful, ever ready to find solutions to conflicts and a middle ground for opposed opinions |Hirschman, 1982: 1465^.

On the other hand, there is no doubt that Smith stressed serious limitations of the self-regulating capacity of the market system. As a matter of fact, he was deeply distrustful of the merchants and manufacturers whom, in The Wealth of Nations, he blamed for being exclusively geared towards advancing their own short-term selfish interests in whichever way possible. His conclusion was that 'the only guarantee that commercial relations will not shred the moral fabric of society is the creation of an institutional framework which sets bounds upon their self-interested action' |McNally, 1988: 193^. This political task -- the constitution and preservation of the moral basis of society -- should be undertaken by a select group of virtuous men capable of following the moral ideals of conscience and embodying the civic habits necessary to political stability |ibid.: 192-208^. In contrast to Hume, Smith believed that there are natural standards of human conduct which are not simply the result of human convention |Young, 1992: 76^. For him, therefore, the social order of the market must be 'fabricated' within the body politic so as to lay down the rules and set up the institutions susceptible of directing self-centred economic appetites into socially desirable channels. Also worth noting is the fact that Smith stressed the role of the Church in producing and strengthening morality |Colclough, 1991^.

Insights from game theory: There is a clear affiliation between most contemporary authors writing on the subject and either of the above two strands of eighteenth-century political thought. Thus, on the one hand, Hayek |1948; 1979^ -- for whom rules of good conduct emerge naturally and get reflected in evolving common law -- neatly belongs to Hume's descent, like many other scholars more or less closely related to him, such as Nelson and Winter |1982^, Gauthier |1986^, Mueller |1986^, Ellickson |1991^, McKinnon |1992^ and Murrell |1992^. Perhaps closest to Hume is the philosopher Gauthier who asserts that an individual 'reasoning from nonmoral premises would accept the constraints of morality on his choices' |Gauthier, 1986: 5^.(31) On the other hand, the sociologist Parsons -- for whom a society is 'prior to and regulates utilitarian contracts between individuals' |Mayhew,
[Long Display] (All, All UC, UCB, UCD, UCI, UCLA, UCR, UCSB, UCSC, UCSD, NRLF, SRLF, Non-UC, CSL, STAN, Greater Bay Area, Northern California, Greater Los Angeles, San Diego/Inland Empire, Central Valley)