Dyadic Processes of Disclosure andReciprocity in Bargaining with Communication

We offer a study revealing the mechanisms through which communication helps actual bargaining behavior outperform economic predictions. The possibility of individually strategic behavior in the presence of private information leads to game-theoretic predictions of less than full efficiency. We present a one-stage, simultaneous offers bargaining game in which buyers and sellers have independent, privately held valuations for the item being sold (i.e. a bilateral auction with two-sided private information). In three communication treatments, parties are: (a) allowed face-to-face communication prior to submitting offers; (b) allowed written communication prior to submitting offers; or (c) allowed no-communication prior to submitting offers. When parties are allowed pre-play communication, we find nearly full efficiency (98%). We examine two systematically predictable aspects of dyadic interaction—disclosure and reciprocity—to explain how negotiators achieve this efficiency. Copyright # 2002 John Wiley & Sons, Ltd. key words negotiation; bargaining; face-to-face; written communication; communication; reciprocity; coordination; honesty; bilateral auctions Both economic and psychological research, as well as casual observation, provide ample evidence that negotiators often fail to behave in the manner envisioned by game-theoretic analyses of bargaining. Much of the existing research attempts to contradict the rationality assumption by providing evidence that negotiators systematically engage in faulty cognitive reasoning and, as a result, fail to reach optimal outcomes (Arrow et al., 1995; Bazerman, 2002; Kahneman & Tversky, 1993). This paper continues in the tradition of uncovering how actual negotiator behavior differs from existing game-theoretic models, but rather than concentrating on sub-optimal behavior, we are concerned with negotiated outcomes that are more efficient than equilibrium outcomes. And rather than focusing on features of individual cognition, we examine systematically predictable aspects of dyadic interaction to help account for situations in which negotiators outperform existing game-theoretic predictions. Copyright # 2002 John Wiley & Sons, Ltd. * Correspondence to: Kathleen L. McGinn (formerly Valley), Harvard Business School, 181 Baker Library, Soldiers Field Road, Boston, MA 02163, USA. E-mail: [email protected] Contract/grant sponsor: NSF. Contract/grant numbers: SES-9210298; PYI-9157447. In many games with private information (any game in which not all of the payoff information is public), equilibrium models predict inefficiencies. Given the presence of private information and less than perfect alignment of interests across the players, models of rational bargaining behavior under standard gametheoretic assumptions predict failures to trade in some instances, despite the existence of a positive bargaining zone (Crawford & Sobel, 1982; Chatterjee & Samuelson, 1983; Farrell & Gibbons, 1989; Matthews & Postlewaite, 1989). The intuition behind the inefficient equilibria is that players holding private information will misrepresent their information in order to maximize their own payoffs. Furthermore, standard game theory predicts misrepresentation, and its subsequent inefficiencies, will occur even when communication between the players is permitted (Myerson & Satterthwaite, 1983). We study a bilateral bargaining game in which a buyer and a seller receive private valuations from a single uniform distribution, have a period for thought and/or communication, and then privately submit bidding and asking prices for the valued item. The terms of trade are determined solely by the bid and ask. If the privately submitted bid is greater than the privately submitted ask, trade occurs at the midpoint. Myerson and Satterthwaite (1983) show that, in this game, there is no equilibrium solution that predicts fully efficient trade. Specifically, the equilibrium that maximizes expected gains from trade involves linear strategies that result in impasses in a substantial portion of cases in which a positive bargaining zone exists (Chatterjee & Samuelson, 1983). In a linear strategy, each party bids his or her true value adjusted by a constant, i.e. they ‘shade’ their bids. The adjustment serves to increase profit if a deal is struck, and only risks the loss of small value trades, thereby increasing expected value over all potential values of the item being traded. Myerson and Satterthwaite’s analysis concludes that even if the players were given the opportunity to communicate prior to play, there is no equilibrium of the enlarged game that produces higher expected gains from trade than Chaterjee and Samuelson’s linear equilibrium. We study precisely this bilateral bargaining game, and include two treatments that allow communication between players prior to the simultaneous, private submission of bids and asks. Our results show outcomes that dominate the linear equilibrium and attain nearly full efficiency (i.e. trade if and only if the buyer’s valuation is greater than or equal to the seller’s valuation) when communication is allowed. We find that players reach pareto superior outcomes through one or more of three dyadic behaviors: (1) coordination on a single price; (2) mutual bidding of values; and (3) mutual revelation of values. Our thesis is that strictly costless, non-verifiable and non-binding talk materially affects the process and outcome of negotiations. Negotiating parties often create interactions in which talk is treated as both verifiable and binding, though there are no within-game structures in place to either bind or verify. As in other social interactions, negotiations are guided by contextually driven norms and practices of interaction (McGinn & Keros, forthcoming). In our results and discussion, we explore how two practices of interaction observed in our study—disclosure and reciprocity—operate in opposition to economic assumptions of ‘cheap talk’. These practices drive outcomes that are substantially more efficient than game-theoretic equilibrium outcomes. COMMUNICATION EFFECTS IN BARGAINING GAMES This is not the first paper to present evidence that communication can produce outcomes superior to equilibrium predictions. Our contribution is an explanation of how norms and practices of communication trigger highly efficient performance in bargaining with private information. Research in the social psychological tradition has found that when communication is allowed in bargaining games with complete information (e.g. dilemma games or coordination games), negotiators systematically reach outcomes that are superior This game is commonly referred to as a double (or bilateral) auction with two-sided private information. 18 Journal of Behavioral Decision Making Copyright # 2002 John Wiley & Sons, Ltd. Journal of Behavioral Decision Making, 16: 17–34 (2003) to economic predictions, even in one-shot, anonymous situations (Orbell, van de Kragt, & Dawes, 1988). Economic research has reached similar conclusions. In an early review, Hoffman and Spitzer (1982) concluded that face-to-face bargaining increased the likelihood of equal splits. Crawford (1990), though he explicitly excluded face-to-face communication in his review, argued that communication affects expectations and enhances coordination. In a meta-analysis of 57 different studies of prisoners’ dilemma and social dilemma games across disciplines, Sally (1995) found that face-to-face communication increased the likelihood of mutual cooperation by 43%, even after controlling for the loss of anonymity and expectations of future interaction. Analogous evidence on the beneficial effects of communication has been reported in games with private information. In a winner’s curse game, where the buyer’s value for a good is yoked to the seller’s value but only the seller has valuation information, Valley, Moag, and Bazerman (1998) found that face-to-face communication during the bidding process enhanced efficiency over that realized with telephone or written communication. While the game-theoretic equilibrium of the winner’s curse game is impasse, Valley et al. (1998) found that over two-thirds of the dyads involved in face-to-face bargaining agreed to trades with positive returns to both parties. Relative to telephone and text-only interaction, face-to-face communication increased the likelihood of honest information exchange, even when such disclosure risked individual maximization of returns. Similar efficiency-enhancing effects for communication have been found in games where both sides hold private information. Raiffa (1982) found 97% efficiency in face-to-face bilateral bargaining, relative to roughly 67% efficiency when bargaining took place through a double auction (the simultaneous submission of bids and asks, with an agreement to split the difference between the bid and the ask whenever the bid was greater than the ask). When Radner and Schotter (1989) allowed agreements to be reached through faceto-face bargaining rather than through a double auction, bargainers captured 99% of all available trades. Radner and Schotter do not investigate the sources of this efficiency, but speculate that the answer lies in social aspects of the unstructured bargaining game. Kagel and Roth (1995) in their Handbook of Experimental Economics offer two hypotheses for the heightened efficiency of bargaining with face-to-face communication: heightened social utility and supplementary channels providing additional information. Valley et al. (2002) show similar efficiencies (94%) when face-to-face communication is allowed prior to determining price through a double auction, suggesting the answer lies not in unstructured bargaining relative to a bidding mechanism, but in the social aspects of communication itself. They provide evidence that the high rates of efficiency are due to enhanced coordination and lower rates of misrepresentation in face-toface interaction, relative to written or no communication. As in the studies of games where only one side holds private information, honest disclosure appears to be a key to the efficiency-enhancing effects of communication in games with two-sided private information. From a game-theoretic viewpoint, communication during bargaining with private information creates a signaling game (Spence, 1973), regardless of the medium in which the communication takes place. In a signaling game where communication is costless, non-verifiable, and non-binding, the communication is considered to be ‘cheap talk’ (Crawford & Sobel, 1982). ‘Non-verifiable’ means the receiving party cannot verify that the message is true. Verbal claims about one’s privately held valuation are non-verifiable in this sense (given the usual implicit assumption in economics that lies cannot be detected), whereas claims about publicly available facts (such as the gross earnings of a publicly held company) would be verifiable. ‘Nonbinding’ means that, if the message contains a promise about future action, the receiver cannot force the sender to carry out this action. A message is non-binding if it is not possible to enforce contracts governing any of the actions under discussion. Verbal claims of willingness to take the good now and pay for it at a later date are non-binding in this sense (‘the check’s in the mail’), while an agreement of payment before delivery would be binding. Cheap talk can provide useful information to the receiver when three conditions are met: the sender’s preferences vary over receivers with different evaluations; the receiver’s preferences vary over the senders with different evaluations; and the interests of the parties are not in complete conflict (Gibbons, K. L. McGinn et al. Dyadic Processes in Bargaining 19 Copyright # 2002 John Wiley & Sons, Ltd. Journal of Behavioral Decision Making, 16: 17–34 (2003) 1992). But cheap talk cannot entirely solve the problem of incomplete information unless the interests of the parties are perfectly aligned (Crawford & Sobel, 1982). In many bargaining contexts, however, strictly costless, non-verifiable and non-binding talk is not ‘cheap’ (in the colloquial sense of low quality). Communication across bargainers materially affects the process and outcome of negotiations because parties treat the information provided in the communications as both verifiable and binding. Rather than relying on within-game structures in place to either bind or verify, they rely on contextually driven norms of honesty and disclosure in communication. In this investigation, we study the bargaining game with bilateral private information, as described above. We introduce communication in the pre-bid period and predict that this communication will resolve the challenge of reaching an agreement, even in the presence of a restricted bargaining zone. Unlike Raiffa (1982) and Radner and Schotter (1989), but similar to Valley et al. (2002), we do not allow completely unstructured bargaining—we restrict communication to the pre-bid period, while the terms of trade are determined solely through the double auction that occurs subsequent to the communication period. We examine the dyadic behaviors observed when communication is allowed and point to the principles of disclosure and reciprocity underlying these behaviors, in an attempt to explain how communication triggers highly efficient performance in bargaining with private information.