Manage. Decis. Econ. 25: 41–54 (2004)
PRINCIPLES OF EVOLUTIONARY PSYCHOLOGY
For instance, one of the entities that we know for sure did not exist in the EEA is television. The fundamental principles of EP would therefore imply that humans have difficulty recognizing and dealing with TV. This indeed appears to be the case. People who watch certain types of TV shows are more satisfied with their friendships, just like they are if they have more friends or spend more time socializing with them in real life. It appears that the human brain has difficulty distinguishing between real friends and imaginary ones they see on TV, because it did not exist in the EEA (Kanazawa, 2002). It is this fundamental observation, that our brain and its psychological mechanisms are strongly biased to view and respond to the environment as if it were still the EEA, which leads to the Savanna Principle.
It is true, as critics of EP often point out, that the EEA, tens of thousands of years past, is not directly observable. We can make inferences about it, based both on archeological records and ethnography of contemporary hunter-gatherer societies, but it is unlikely that we will ever know all the details of the EEA. It is therefore impossible for us to draw all the implications of the above observation for our current social behavior. However, there are certain things about our ancestral life in the EEA that we know reasonably well. We know that our ancestors lived in small bands not exceeding 200 individuals; they did not live in a metropolis where everybody can be anonymous. We know that all communications between people in the EEA were direct and face-to-face; they did not have telephones, computers or even writing that allowed them to communicate without facing each other. It is my suggestion in this paper that these few facts that we know about the EEA are sufficient to use the Savanna Principle to figure out which hypotheses about human behavior are likely to fail and why.
In PDGs [Prisoner’s Dilemma games], defection strictly dominates cooperation. No matter what the other player chooses, each player is better off defecting than cooperating. Mutual defection, which is collectively the worst outcome, thus becomes a Nash equilibrium, from which neither player has any incentive to deviate unilaterally (Nash, 1951). In both one-shot and finitely repeated PDGs, mutual defection is the only Nash equilibrium. Noncooperative game theory therefore predicts that all rational players choose to defect, and mutual defection is the only possible collective outcome. Since cheap talk (nonenforceable threats and promises) does not alter the players’ payoffs and payoffs are the only determinants of behavior, the theory also predicts that communication among the players has no effect on their choices.
That is not how all human subjects behave, however. In a comprehensive review of the experimental literature on PDGs, Sally (1995) concludes that roughly half (47.4%) of all subjects in 130 different experiments published in 37 studies make the ‘irrational’ choice to cooperate. Further, of all the factors considered by experimentalists in 35 years, cheap talk probably has the largest positive effect on cooperation. Experimental subjects are significantly more likely to cooperate when they can communicate with each other before and during the experiment, and exchange promises and threats with each other, even though such promises and threats are not enforceable by the rules of the game.
There are two anomalies here. Both what Sally (1995, pp. 70–72) calls the ‘strong self-interest hypothesis’ (No rational actor will ever cooperate in PDGs) and ‘weak self-interest hypothesis’ (Only factors that change the players’ objective payoffs at the margin affect their choice) derived from noncooperative game theory fail. A large number of subjects do choose to cooperate, and cheap talk does have a positive effect on cooperation. What is wrong here?
One of the key assumptions in noncooperative game theory is that the two (or more) players of PDGs are completely anonymous. Experimentalists go to great lengths to make sure that their subjects do not meet before, during and after the experiment. The experimental design guarantees the anonymity of the subjects. It is this complete anonymity, and thus the impossibility of knowing future interactions, that partly lead to the prediction that defection is the only rational choice and that cheap talk has no effect on cooperation. For if the subjects knew the identities of each other, then there will be other considerations besides the payoffs from the game. Subjects may fear retaliation (physical or otherwise) from the other players when they defect on them or when they do not honor their promises to cooperate (however non-enforceable within the rules of the experiment). They may fear that defection or breaking promises might ruin actual or potential friendship or acquaintanceship, that it might engender ‘bad feelings’ between them. They may fear that their reputation might be ruined if they are perceived as selfish defectors or as someone who doesn’t keep their promises. Complete anonymity between subjects guarantees that none of these considerations are relevant for their utility calculations.
However, it is likely that no such complete anonymity existed in the EEA. As our ancestor during the Pleistocene epoch, you lived in a small band of 50 to 200 individuals, where everybody knew everybody else. Further, the only way for you to communicate and interact with others in the EEA was to face them directly. There were no anonymous means to communicate or interact. So whatever choice you made in interpersonal relations was known at least to your partner, who knew who you were, if not to everyone else in the band. If you defected on Og or didn’t keep your promise with him, he would likely tell everyone in the band that you were an untrustworthy cheater, probably after he beat you up first. Complete anonymity, which is an integral assumption in noncooperative game theory, probably did not exist in the EEA, and the human brain, biased to perceive the environment as if it were still the EEA, cannot quite comprehend such a thing.
Another integral assumption in noncooperative game theoretic prediction of mutual defection in one-shot PDGs is noniteration. Rules of the game in one-shot PDGs stipulate that the two or more players meet only once to make their choices (either cooperation or defection) and they will never meet again. (Complete anonymity helps guarantee it.) Noniteration is integral to the prediction of mutual defection, because Axelrod (1984) has demonstrated that mutual cooperation becomes rational and a sustainable Nash equilibrium when the same players meet repeatedly and indefinitely, and they play Tit-for-Tat or other contingent strategies.
Once again, this assumption of noniteration probably did not hold in the EEA. As our ancestor, your social world was limited to others in your band and possibly in the neighboring bands. For the most part you interacted with the same people your entire life (although occasional outmigration might have been possible). Just as you could not have complete anonymity in the EEA, you could not have a ‘one-shot’ interaction with anyone with no possibility of iteration. Every interaction in the EEA was likely an indefinitely repeated game. The Savanna Principle explains why a hypothesis based on the assumptions of complete anonymity and noniteration – the prediction that everyone will defect in PDGs – fails. If there were no complete anonymity or impossibility of future interactions, then cooperation and honoring promises you make in cheap talk suddenly become rational.
In a recent article, Kiyonari et al. (2000) present an alternative evolutionary psychological explanation for cooperation in one-shot PDGs. They argue that individuals possess ‘social exchange heuristic’, which compels them to play PDGs as if they were Assurance games. In PDGs, unilateral defection is preferable to mutual cooperation; in Assurance games, mutual cooperation is preferable to unilateral defection. Individuals playing Assurance games, unlike those playing the PDGs, are therefore motivated to cooperate as long as the other player also cooperates. Now what transforms the PDGs into Assurance games in the minds of many individuals? Infinite iteration. It is only with the infinite iteration and the use of contingent strategies such as Tit-for-Tat that PDGs are transformed into Assurance games, and mutual cooperation becomes preferable to unilateral defection. Kiyonari et al. (2000) argue that individuals have the social exchange heuristic which assumes that all games are infinitely iterated (and are therefore Assurance games rather than PDGs), because that was the nature of social exchange in the EEA. Their theory of social exchange heuristic and their experimental data are therefore perfectly consistent with the Savanna Principle.
Collective action purports to provide public goods. Unlike private goods, public goods, once provided, are nonexcludable (both contributors and noncontributors to their production can consume them) and have jointness of supply (consumption by some does not decrease the amount left for others to consume). Rational actors therefore have no incentive to contribute voluntarily to the provision of public goods. They can free ride on others’ contributions and consume the public goods when they are provided (since free riders cannot be excluded from consumption). If everyone makes the rational decision, however, no one will contribute toward the provision of public goods, and they will never be produced (Olson, 1965). Public choice theory predicts that all rational actors free ride, and hence public goods will never be provided. This is the essence of the collective action problem. Benefits of collective bargaining and industrial action are examples of public goods relevant to managerial and organizational economics.
Contrary to theory, however, public goods are routinely provided, both in laboratory experiments and in natural settings. While the level of provision is often less than optimal, subjects do contribute their private resources toward the production of public goods in laboratory experiments (Ostrom, 1998). In natural settings, examples of successful collective action abound, from worker strikes to political protests to consumer boycotts to nationalist movements. Why do individuals participate in such collective action when the benefit (be it higher wages or political change or safe consumer goods or ethnic independence) cannot be excluded from free riders who do not participate?
An integral assumption in public choice theory is that the collective action is large, involving thousands or millions of people. The large scale of the collective action leads to two conditions: anonymity of individual choices and negligibility of each actor’s contribution. Because it involves a large number of actors, each actor’s choice to cooperate or defect is not known to others (this is institutionally guaranteed in some collective actions, like voting in democratic societies), and each actor’s contribution makes a negligible difference to the collective outcome. Nobody knows whether you contributed or defected, and your contribution or defection makes very little difference to whether the collective action succeeds or fails.
It is likely that any collective action that took place in the EEA was small in scale (Rubin, 2001b). Thus neither consequence of large collective action (anonymity and negligibility of individual contribution) probably existed in the EEA. Whether you participated in an coordinated effort to hunt big game or collective childcare arrangement was immediately known to everyone else in the band. And your contribution made a significant difference to the collective outcome, when there were only dozens of actors at most. Whether or not the hunt became successful, and as a consequence you and your family ate animal protein that day, could crucially depend on how much effort you, personally, put into the coordinated hunting. It becomes rational to contribute to collective action when your individual share of the public goods approaches or exceeds your individual contribution toward their production, especially when others with whom you spend your entire life know whether or not you contributed. The human brain, adapted to the EEA, may still respond to many instances of collective action, however large in scale, as if they involved only dozens of people, and thus it might be rational to contribute (Rubin, 2001b).
Counterexample: Network Exchange Theory
So far I have discussed how the Savanna Principle can explain why some hypotheses about human behavior fail by not taking the EEA into account, and used noncooperative game theory and public choice theory as examples. Both theories are often tested (and at least partially disconfirmed) in laboratory experiments. Despite rigorous controls that laboratory experiments allow, hypotheses derived from these theories often prove relatively unsuccessful, consistent with the Savanna Principle.
Another theoretical perspective, network exchange theory in sociology, provides a sharp contrast. Network exchange theory originates with the work of Emerson (1962, 1972a, 1972b), and explains actors’ behavior in terms of their power inherent in their positions (nodes) in exchange networks. Holding the value of resources constant, actors have more power (and can thus bring about more favorable outcomes for themselves) if they have more exchange partners who themselves have fewer alternatives. Molm (1997) and Willer (1999) provide excellent reviews of network exchange theory. Its most fruitful application to managerial and organizational theory is Burt’s (1992, 2000) structural holes theory, which predicts that those in corporate organizations who occupy structural holes (network nodes that are connected to other nodes that are themselves not connected) have social capital because they can function as information brokers within the organization. His data show that employees and managers who occupy such structural holes tend to be promoted faster.
Unlike noncooperative game theory and public choice theory, hypotheses derived from network exchange theory are usually confirmed by experimental data, often very precisely, down to the decimal point. While there are minor differences in various theories within this perspective (Skvoretz and Willer, 1993), hypotheses derived from all of them are quite successful by the social science standards. Due largely to its formal models and standardized experimental procedures, network exchange theory is probably one of the very few fields in social sciences that qualifies as what Collins (1994) calls ‘high-consensus, rapid-discovery’ science, reminiscent of natural sciences.
Network exchange theory is similar to noncooperative game theory and public choice theory in that it conceives of the actor as rational, purposive and self-interested. It is also similar to them in its degree of formalization and its frequent use of laboratory experiments for testing hypotheses. If network exchange theory is similar to noncooperative game theory and public choice theory in its assumptions about the actor, deductive mathematical models, and preferred method of testing, why is it so much more empirically successful than them? From the perspective of the Savanna Principle, the answer may lie in the extent to which the scope conditions and assumptions of network exchange theory are consistent with conditions that prevailed in the EEA.
The single most important predictor of behavior in network exchange theory is power. Actors who find themselves in certain structural locations in networks have more power than those in other structural locations. When more powerful actors interact with less powerful actors, the former can obtain outcomes that are more favorable to them than to the latter. Actors’ structural power is inversely determined by their dependence on others, and dependence is in turn determined inversely by the number of potential exchange partners and directly by the value they place on the resources that their exchange partners possess (Emerson, 1962). Actors are more dependent on their exchange partners to the extent that they have fewer alternative exchange partners (fewer network ties to others) and that they place greater value on the resources that their exchange partners hold. The more dependent the actors are on others, the less power they have over them. These are the basic principles of network exchange theory.
Note that none of these assumptions appear inconsistent with the EEA. Our ancestors most probably exchanged resources with each other. Some had more exchange partners (more network ties to others) than others, and as a consequence were less dependent on and exercised more power over others. Unlike noncooperative game theory or public choice theory, hypotheses derived from network exchange theory do not presume complete anonymity among actors or a large group size. If they required complete anonymity among actors, then they would have to predict that actors do not exercise power over others in face-to-face interactions. They would not predict, for instance, that the most popular girl in high school (who can get many dates) can get more out of her dates than the least popular girl (who can get few dates), or that a job candidate with many offers can extract a more favorable contract than one with few offers. However, these predictions are perfectly consistent with network exchange theory. None of the French or American corporate managers Burt (2000) studies are anonymous to each other within their firms.
One of the network exchange theories that Skvoretz and Willer (1993) test is Bienenstock and Bonacich’s (1992) core theory based on cooperative game theory. In stark contrast to noncooperative game theory, which explains strategic choices of individuals, cooperative game theory (Kahan and Rapoport, 1984) models coalition formation. Within cooperative game theory, all agreements are enforceable and anonymity is not necessary, reflecting the conditions that probably prevailed in the EEA more accurately than noncooperative game theory. The Savanna Principle can therefore explain why cooperative game theory makes more accurate empirical predictions about human behavior than noncooperative game theory.
None of the integral scope conditions and assumptions of network exchange theory are inconsistent with what prevailed in the EEA. Hypotheses derived from network exchange theory would probably have been successful in the EEA, while those derived from noncooperative game theory and public choice theory would not have been. Perhaps the strongest indication for this is that network exchange theory has been tested and supported with nonhuman primate species (Maryanski, 1987; Maryanski and Ishii-Kuntz, 1991). It is therefore likely that scope conditions and assumptions of network exchange theory held true even before the EEA in the evolutionary history of primates, before our ancestors were human.
Counterexample: Competitive Price Theory in Double Auction Markets
Another theory that is even more empirically successful than network exchange theory is competitive price theory tested in double auction markets in experimental economics (Smith, 1962, 1964). In a typical double auction experimental market, there are several sellers and buyers. Each seller has a certain units of commodity to sell on the market, and each buyer has a certain units of the same commodity to buy. The cost of each unit of the commodity can vary between the sellers, and the value of each unit of the commodity can vary between the buyers. When an auction period begins, each seller posts an ‘ask’ (asking price), and each buyer posts a bid. Sellers successively lower their asks, and buyers successively raise their bids, until there’s a match between an ask from a seller and a bid from a buyer, at which point they enter a binding contract. The auction period ends either when the sellers sell all the units they want to sell, or the buyers buy all the units they want to buy.
Double auction markets function so well that their mean efficiency in various experiments range from 95 to 100% (Davis and Holt, 1993, p. 136, Table 3.4; Holt, 1995, p. 371, Table 5.2). In other words, 95–100% of theoretically possible surplus is actually extracted by buyers and sellers in double auction experimental markets. In fact, double auction markets are so efficient that they serve as benchmarks against which the performance of all other market institutions is evaluated.
Why are double auction markets so efficient? Competitive price theory, which these experimental markets are designed to test, rests only on a few fundamental economic concepts: Demand, supply, value, and cost. These four parameters are sufficient to compute the competitive price for any market. None of these concepts were absent in the EEA, and thus violate the Savanna Principle. Even in the EEA, some goods were in greater demand (food) than others (flowers). Some goods were in greater supply (berries in season) than others (berries out of season). Some goods had inherently greater value to people (sharp spears) than others (dull spears). Some goods were inherently more costly to produce (meat of large game) than others (meat of small game). In their economic exchange, our ancestors would have demanded or offered more for inherently more valuable or costly goods than for inherently less valuable or costly goods, and they would have demanded or offered more for goods in greater demand or in shorter supply. Our ancestors would have made as good subjects for double auction experimental markets (once they overcome the language barrier) as sophomores at California Institute of Technology or the University of Arizona.
It is important to note that, while double auction markets in experimental economics (like experiments testing network exchange theory) are often conducted via computers and thus participants in these experiments remain anonymous, neither computerized (indirect, non-face-to-face) communication nor anonymity is an integral assumption of competitive price theory ( just like they are not for network exchange theory). In fact, early double auction markets were conducted face-to-face (in what is now known as the ‘oral’ double auctions), and they were slightly more efficient than the computerized markets with the same parameters (Williams, 1980). Prices are also more variable and volatile in computerized double auction markets than in their oral counterparts (Davis and Holt, 1993, pp. 135–141; Williams, 1980).
Computerized and oral double auction markets function more or less the same because the equilibrium price predicted by competitive price theory depends only on demand, supply, value and cost. If computerized communication and anonymity were integral to auction markets, then one would have to predict that computerized, anonymous auctions like eBay would perform differently than face-to-face, nonanonymous auctions like Sotheby’s. Disregarding for the moment that eBay typically deals with an entirely different class of commodities than Sotheby’s, competitive price theory nonetheless predicts that both auction markets would reach the same equilibrium price equally efficiently. 2
2. While anonymity is not essential for auction markets, reputation is. Sotheby’s is not likely to invite just anyone off the street into its auction gallery and accept her personal check, and participants in eBay are not likely to transact with someone with a reputation for cheating. That is why eBay provides members’ reputation scores on its site, while preserving their anonymity through computerized transaction via anonymous handles. The genius behind eBay’s success is eliminating what is not necessary for efficient auction (face-to-face transactions) while preserving what is (reputation).
This is in stark contrast to noncooperative game theory and public choice theory, for which anonymity (created by their computerized experiments) is an integral theoretical assumption. Participants in face-to-face PDGs are predicted to behave differently than those in anonymous PDGs. And they do. Participants in face-to-face PDGs and similar games often experience extreme rage toward defectors and threaten them with physical violence (Bonacich, 1976, pp. 206–208; Ostrom et al., 1992). That is why anonymity is necessary. It is unlikely that participants in Sotherby’s auction would threaten someone who just outbid them with physical violence if the auction was otherwise fair. The Savanna Principle predicts empirical failure only when the theory’s integral scope conditions or assumptions are inconsistent with the EEA. It can therefore explain why network exchange theory and competitive price theory perform so much better empirically than noncooperative game theory and public choice theory.
Finally, a critic might argue that an alternative explanation for the success of network exchange theory and competitive price theory is the ‘disciplinary’ power of competition. 3 In situations modeled by these theories, unlike those modeled by noncooperative game theory and public choice theory, some individuals are routinely excluded from exchange. An actor who offers less to a potential exchange partner than someone else will not be chosen for the exchange. A buyer who bids less than another buyer will not win the bid. In order not to be excluded, individuals must modify their behavior continuously until it nears efficiency predicted by theory; in other words, competition, and the possibility of exclusion, ‘disciplines’ the actors’ behavior until it is optimal, regardless of their cognitive processes and what their brain can or cannot recognize (as the Savanna Principle posits).
While this alternative perspective can explain why network exchange theory and competitive price theory succeed in predicting human behavior precisely, it cannot explain why noncooperative game theory and public choice theory fail. In contrast, the Savanna Principle can simultaneously explain why (and which) theory succeeds empirically, and why (and which) one fails. The Savanna Principle is also consistent with the fact that it doesn’t seem to take much ‘discipline’ for inexperienced participants to reach equilibrium prices. In Smith and William’s (1983) experiment, for instance, inexperienced participants reach 94.8% efficiency only after two trading periods and 100% efficiency after five. It is highly doubtful that inexperienced participants in experimental markets can reach efficiency so quickly unless their brain is already equipped with the concepts of demand, supply, value and cost.
Some laboratory experiments have demonstrated that players are more likely to cooperate with each other when they can see each other in PDGs, even though they cannot otherwise communicate (Wichman, 1970). Sally’s (1995) meta-analysis shows that visibility has a strong positive effect on cooperation at least for repeated games. Once again, a factor that does not alter players’ payoffs influence the rates of cooperation, and the positive effect of visibility is therefore a mystery. Since noncooperative game theory predicts mutual defection in PDGs, this means that the hypothesis fails even more, and the Savanna Principle holds even stronger, when the players can see each other.
Computerized Cheap Talk
… When the experimental subjects have cheap talk by exchanging messages through computer terminals, without facing each other directly, the increase in cooperation is not as great as if the communication was face-to-face (Isaac and Walker, 1988; Sell and Wilson, 1991) or is sometimes completely nonexistent (Palfrey and Rosenthal, 1988). In other words, one of the most robust (if unexplainable within noncooperative game theory) findings in the PDG literature does not occur (or occurs to a far lesser extent) when the communication and cheap talk happen via computers. Since noncooperative game theory predicts mutual defection throughout, this means that the Savanna Principle (stated negatively), predicting the failure of the hypotheses, holds less with computerized cheap talk.
RECOMMENDATIONS FOR BUSINESS AND MANAGEMENT
1. Do not treat men and women identically and interchangeably. An incredibly ingenious recent experiment (Kurzban et al. 2001) convincingly demonstrates that, unlike race categories, sex and age categories are genetically hardwired in the human brain. This finding makes perfect sense from the perspective of the Savanna Principle because the sex and age distinctions have always existed in identical forms throughout the human evolutionary history, while what constitutes an ingroup (whose members are to be favored) and an outgroup (whose members are to be disfavored) depended on what constituted a deme (an endogamous group) in the local society (Whitmeyer, 1997). Thus, while ethnocentrism is probably hardwired, what constitutes an ethnic group is not.
It should therefore be very difficult, if not virtually impossible, for the human brain to treat men and women identically and interchangeably, as difficult as it is for us to believe that sugar is distasteful and rotten meat is delicious. The current law that requires that employers and employees treat men and women identically and interchangeably goes against the core of human nature, and, just like anything else that goes against human nature, it is likely to fail. Besides, it makes no sense to treat men and women identically and interchangeably because ample empirical evidence incontrovertibly demonstrates that men and women are inherently and biologically different, and that the sex differences in behavior and outcomes probably result from such biological differences in preferences and dispositions, not from employer discrimination (Browne, 2002).
2. Do not treat the young and the old identically and interchangeably. Because age categories, like the sex categories, are genetically hardwired and cannot therefore be ‘erased’ (Kurzban et al., 2001), any law that requires that employers and employees treat the young and the old identically and interchangeably also goes against the core of human nature and is bound to fail. The obverse of the Savanna Principle would therefore recommend that employers not ignore the age differences and that they not require their employees to do so either.
Once again, the identical and interchangeable treatment of the young and the old would make no scientific (and thus economic and management) sense since age, like sex, is an important determinant of preferences, predispositions and behavior. Just as ‘sex discrimination’ is not necessary to explain sex differences in behavior and outcomes, ‘age discrimination’ is not necessary to explain age differences in behavior and outcomes. Evidence shows that young, unmarried men are responsible for an overwhelming majority of productivity in such widely divergent fields as music, art, literature and science (Kanazawa, 2000; Miller, 1999). Young, unmarried men are intensely competitive and driven to attain a higher status whenever and wherever they find a status hierarchy, no matter now trivial, meaningless or even illegal (Wilson and Daly, 1985). It appears that employers and managers ought to take advantage of such extra energy in the pursuit of corporate goals.
The extreme competitiveness of young men, which leads to their enormous productivity, is a double-edged sword, however. 4 Their competitiveness manifests itself in their risk-taking behavior (Wilson and Daly, 1985), which is often essential for success in new economic ventures but can be fatal for more routine management decisions with much at stake. For the same evolutionary developmental reasons that make young men competitive and risk-taking, older men are more conservative and risk-averse, because older men who became our ancestors and from whom we inherit our psychological mechanisms, had achieved high statuses by their late adulthood, which they could potentially lose if they continued to be competitive and take risks (Kanazawa and Still, 2000; Kanazawa, 2001, pp. 1151–1153). Thus, while young men are better suited for leading new ventures into uncertain territories, older men and women, who are even more risk-averse than older men (Campbell, 2002), are probably better for routine decision-making.