Tuesday, July 03, 2012

The Weakness of Strong Reciprocity

"How selfish soever man may be supposed, there are evidently some principles in his nature, which interest him in the fortune of others, and render their happiness necessary to him, though he derives nothing from it except the pleasure of seeing it."

That's the first sentence of Adam Smith's Theory of Moral Sentiments.  It's the first sentence of the first chapter, which is on "sympathy" as "our fellow-feeling with any passion whatever"--the fundamental mechanism for Smith's explanation of all human morality.  This first sentence announces Smith's opposition to those like Thomas Hobbes and Bernard Mandeville who apparently thought that human beings could be explained as purely selfish beings,  because Smith believes that human beings are not just selfish but also social animals.

In The Descent of Man, Darwin cites this chapter in his own account of sympathy as the natural ground for human sociality and the moral sense.  And so it seems that Darwinian social theory agrees with Smith that we need to recognize that human conduct is other-regarding as well as self-regarding in its motivations.

Samuel Bowles and Herbert Gintis quote this first sentence of Smith's book as the epigram for the first chapter of their recent book, A Cooperative Species.  In this book as well as other writings, Bowles and Gintis suggest that in criticizing the "self-interest axiom" of the Homo economicus model and affirming "strong reciprocity" as a powerful human motivation shaped by human evolution, they are following in the tradition of Smith and Darwin. 

And yet both Smith and Darwin indicate that human sociality can be largely explained as naturally rooted in kinship and reciprocity.  The filial affections for one's children and other relatives are extended to other members of one's community.  And we can do good to others with the expectation of receiving good in return.  The first corresponds to William Hamilton's "kin selection," the second to Robert Trivers' "reciprocal altruism."  The idea of strong reciprocity as defended by Bowles and Gintis goes beyond this, however, because strong reciprocity is a propensity to cooperate and share with others, even strangers, and to punish those who don't cooperate, even when cooperation and punishment are personally costly to the reciprocator.  Thus, strong reciprocity does not require either ties of kinship or expectations of future reciprocation.

Studying the arguments for strong reciprocity by Bowles and Gintis in A Cooperative Species, alongside the arguments of their critics, I am inclined to conclude that if strong reciprocity exists, it's weak in the sense that it's displayed by only a few individuals in special circumstances,   The problem in the book is that at critical points Bowles and Gintis pass over serious objections without adequately responding to them, and in some cases they don't even acknowledge the objections.  Here I will point to four examples of this.

My first example is their response to some of the experimental research with the Ultimatum Game and the Dictator Game.

In the Ultimatum Game, the experimenter allocates some amount of money (say $10) to one person--the "proposer"--who must offer some proportion of this money to another person--the "responder"--who is free to accept or reject the offer.  If the responder rejects the offer, neither person receives any money.  If the responder accepts the offer, then the money is divided according to the offer of the proposer.  If this is conducted as a one-shot, anonymous game, then the Homo economicus model of human beings as rational egoists would predict that the proposer would offer to share as little as possible--maybe $1--and the responder would accept, because $1 is better than nothing.  But actually, experimenters have reported that most proposers regularly offer between 40% ($4) and 50% ($5) of the money.  When proposers offer less than this, responders generally reject the offer.  This has been interpreted as indicating that people act not purely out of self-interest in maximizing their material gain but out of an other-regarding sense of fairness.  The crucial factor seems to be the willingness of the responder to bear some cost in punishing unfair offers, which is anticipated by the proposer. 

Near the beginning of their book, Bowles and Gintis present this as a refutation of the Homo economicus model and as evidence for strong reciprocity (19-20).  (I suspect that when Bowles and Gintis first formulated the idea of strong reciprocity in 1998, this was primarily an explanation for the experimental outcomes with the Ultimatum Game.)  But then, later in the book, they admit that experiments conducted by Vernon Smith and his colleagues show that strong reciprocity is considerably weakened when the Ultimatum Game is played in a social context more like real life.  Bowles and Gintis write:
Evidence that institutions serve as cues for appropriate behaviors comes from ultimate game experiments with U.S. subjects in which simply naming the game "The Exchange Game," or assigning the role of proposer to those who did well on a current affairs test, resulted in lower offers and a significant reduction in rejections of low offers (Hoffman et al. 1994b). If individuals cared only about their money payoffs, neither manipulation would have changed the game. The fact that significantly less strong reciprocity occurred in the exchange game and the current events test version suggests that social structure affects behavior in ways other than those captured by the money payoffs of the game, in this case by suggesting appropriate behavior (the exchange game) or identifying some individuals as "deserving" (the test manipulation). (34)
The common practice in behavioral game experiments in the Ultimatum Game had been for the experimenter to provide the initial endowment of money to someone randomly chosen as the proposer, who then must decide how to divide the money with the responder. 

Vernon Smith's insight was to realize that this is not what normally happens in real life.  Usually, money is initially perceived as someone's property.  To simulate this, Smith had the participants in the game take a test on current events, and whoever scored the best was chosen to act as the proposer with the initial endowment of $10.  This suggested that the proposer had "earned" the money as his property, and the responder knew this.  In some cases, Smith added another condition: he told the participants that they were engaged in an "exchange game."  In those games where the proposer had earned his money as his property, the proposer offered less than $5 to the responder, and the responder accepted the offer.  In those games where the proposer earned a property right and the game was called an "exchange," the proposer offered even less money to the responder, and the responder accepted, although some responders rejected offers of only $1.

From this, Smith concluded:  "both proposers and responders in ultimatum games take account of the conditions under which rights to act have been conveyed.  In particular, a person with a legitimate right believes he/she can use that right in a more self-regarding manner than when the right is ambiguous, ill defined, or illegitimate, and others (responders, in this case) agree with, or respect, those beliefs of the rights holder" (1998, 14).  So what Bowles and Gintis identify as strong reciprocity is very weak when people are dividing up money that is perceived as belonging to someone.

Oddly enough, Bowles and Gintis are completely silent about the other research reported in the article they cite by Smith and his colleagues.   The Ultimatum Game becomes a Dictator Game when the responder has no right to reject the offer.  This allows the proposer to be selfishly unfair in dividing the money without any fear of punishment by the responder.  But while a purely selfish dictator would be free to take all the money and share none of it, the typical outcome of the Dictator Game is that 60%-80% of the dictators give $1 or more, although dictators tend to give less than they would if they were proposers in an Ultimatum Game. 

Smith noticed, however, that while experimenters enforced anonymity among the players in the Dictator Game, so that dictators would not worry about their reputation, the players' decisions were being recorded by the experimenters, who then would know what the dictators had done.  Smith decided to conduct a series of Dictator Games with a double blind procedure so that the individual decisions of the dictators would be known to themselves but not to the other players or to the experimenters.  When the games were played this way, 64% of the dictators kept all the money for themselves, and 20% of them gave only $1.  From this, Smith concluded: "The appearance of other-regardingness comes from the self-regarding requirements of, and the need for, reciprocity in social exchange.  Take away all social context--no others can know--and we see the naked expression of purely self-regarding behavior" (1998, 16).

This research by Smith and his colleagues suggests that what Bowles and Gintis see as evidence for strong reciprocity is actually evidence of indirect reciprocity in which people are generous because they expect to earn a reputation for generosity that will benefit them in the future.  Under conditions of total anonymity, people tend to express their purely selfish motivations.  Nowhere in their book do Bowles and Gintis respond to this evidence.

But notice that even in the double blind Dictator Game, 16% of the dictators gave away more than $1.  Would Bowles and Gintis say that this shows that at least a small minority of people are strong reciprocators?  Surprisingly, Smith reports that in one of his double blind Dictator Games, one individual dictator gave away $9 and kept only $1 for himself!  Would Bowles and Gintis say that surely at least this one individual was a strong reciprocator?

Or should we wonder whether the benevolent dictator who gave away $9 was an undergraduate student who thought the whole game was a joke?  After all, $10 is such a small amount of money for most American college students that deciding how to divide it up between two individuals might not be a realistic test of stinginess or generosity.  This indicates the problem of whether these experimental games are really valid indicators of how people behave in everyday life.

I'll take up this and other issues in my next post.


Vernon Smith, "The Two Faces of Adam Smith," Southern Economic Journal, 65 (1) (1998): 1-19.

Vernon Smith, Bargaining and Market Behavior: Essays in Experimental Economics (Cambridge: Cambridge University Press, 2000).

No comments: