Deception in Networks: A Laboratory Study

Deception in Networks: A Laboratory Study Rong Rong and Daniel Houser April 2014 Discussion Paper Interdisciplinary Center for Economic Science 4400 University Drive, MSN 1B2, Fairfax, VA 22030 Tel: +1-703-993-4719 Fax: +1-703-993-4851 ICES Website: http://ices.gmu.edu ICES RePEc Archive Online at: http://edirc.repec.org/data/icgmuus.html Deception in Networks: A Laboratory Study Rong Rong*1and Daniel Houser** April, 2014 Abstract: Communication between departments within a firm may include deception. Theory suggests that telling lies in these environments may be strategically optimal if there exists a small difference in monetary incentives (Crawford and Sobel, 1982; Galeotti et al, 2012). We design a laboratory experiment to investigate whether agents with different monetary incentives in a network environment behave according to theoretical predictions. We found that players’ choices are consistent with the theory. That is, most communication within an incentive group is truthful and deception often occurs between subjects from different groups. These results have important implications for intra-organizational conflict management, demonstrating that in order to minimize deceptive communication between departments the firm may need to reduce incentive differences between these groups. JEL classification: D85, D02, C92 Keywords: social networks, deception, strategic information transmission, experiments !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! *Department!of!Economics,!Weber!State!University,!Ogden,!UT,!84408,!Rong:[email protected]!! **Interdisciplinary!Center!for!Economic!Science!(ICES)!and!Department!of!Economics,!George!Mason!University,! Fairfax,!VA!22030,!Houser:[email protected]! We!thank!NSF!Dissertation!Improvement!Award!for!financial!support!of!this!project.!For!helpful!comments!we! thank!our!colleagues!at!ICES,!George!Mason!University!and!Goddard!School!of!Business!and!Economics!at!Weber! State!University,!seminar!participants!at!the!ESA!NorthXAmerican!meeting!(2012),!the!Research!Brown!Bag! Meeting!at!Utah!Valley!University!and!the!Networks!and!Externalities!Meeting!at!Louisiana!State!University.!The! authors!are!of!course!responsible!for!any!errors!in!this!paper.! ! ! I. Introduction Groups with different financial incentives often deceive each other for strategic reasons. Within an organization, for example, people from different departments often manipulate the information they send to each other so that executive decisions will be in their favor. A familiar example occurs when academic departments make hiring decisions. Faculty members in a specific field may withhold important information about certain candidate from faculty of other fields, hoping to raise the priority of hiring a colleague in one’s own field. Not surprisingly, this phenomenon has also been observed in many non-academic organizations, including important business sectors such as high-tech research and development, mass media and health care (Cloke and Goldsmith, 2000; Cowan, 2003; Tobak, 2008, Gupta et al, 1985; Eckmen and Lindlof, 2003; Pirnejad et al, 2008). The negative impacts of deception due to the conflict of interest have been documented widely in the studies of industrial and organizational psychology, as well as management (Colb et al, 1992; Rahim, 2000; Dreu and Galfand, 2007; Conrad and Poole, 2011; Miller, 2011). Economists have studied deception using sender-receiver games. Seminal work by Crawford and Sobel (1982) describes a one sender and one receiver case (also denoted a strategic information transmission game, or cheap talk game)2. In their model, the informed sender sends a “cheap talk” message to an uninformed receiver. Then the receiver, as the only decision maker of the game, would choose the option that determines the payoff for both players. Their model provides the conditions where uninformative messages (“cheap talk”) are the equilibrium outcome of the game. Many other studies have used variations of this model; however, prior to Galeotti et al (2013), players either made decisions as a sender or as a receiver, but never both3. Galeotti et al (2013) investigates N player communication in a network setting where one can send cheap-talk messages to others and also receive messages from others. Despite the complexity, their model generates sharp predictions when the players are divided into two groups. In this “two group model,” players’ payoffs are the same within a group but differ between groups. The model predicts that truth-telling among those with aligned incentives will be greater than where incentives are misaligned. This is very much the case for academic hiring. Assuming the “players” are divided into micro and macro faculty, micro faculty enjoy higher payoff (through research synergy) when the new hire is another micro-economist and vise versa. According to the model, one would expect to see higher level of truthful communication within micro or macro faculty and less so between the two. The goal of this study is to test these predictions of the “two group model” and, in particular, to investigate to what extend will people lie to achieve higher monetary gain in the network senderreceiver experiment. !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 2 !Experimental!studies!support!this!prediction!includes!Dickhaut!et!al!(1995),!Blume!et!al!(1998),!Blume!et!al! (2001),!Cai!and!Wang!(2006)!and!Wang!et!al!(2010).! 3 !An!exception!is!Hagenbach!and!Koessler!(2010).!Their!paper!has!very!similar!setup!of!the!model!comparing!to! Galeotti!et!al!(2013).!It!also!yields!similar!predictions.!We!delay!the!discussion!of!Hagenback!and!Koessler!(2010)!to! the!literature!review!section.! ! ! Our experiment tests the two-group cheap talk model in the lab. We believe a laboratory analysis is ideal for this study. The reason is that in natural environments it can be difficult to identify the causal impact of monetary incentives on truth-telling. In particular, when using non-experimental observations, the empirical correlation between the two may not convey a causal story convincingly: there may be other factors impacting both the incentives one faces as well as one’s communication strategy. As the purpose of this study is to discover how monetary incentive alone impact deception, we randomly assign monetary incentives to each subject. Our main findings are (1) consistent with theory, truth-telling nearly always occurs among those with identical monetary incentives; (2) systematic over-communication occurs between groups with different incentives; and (3) players overly trust messages they receive. To our knowledge, we are the first to provide empirical evidence on behavior in sender-receiver games with multiple senders and multiple receivers4. Despite the many insights gleaned from one-sender-one-receiver cases, extending the strategic information transmission to a group context is important. It provides more accurate description on the types of communication that occurs in multi-group population with divergent preferences. The remainder of this paper is organized as follows: Section 2 briefly reviews some of the related theoretical and experimental literature. Section 3 lays out the theoretical background for our study. Section 4 presents the experimental design and procedures. Section 5 describes the hypothesis and reports experimental results. Section 6 concludes. II. Literature on Deception There are a number of economic theories and experimental tests of sender-receiver environments. We begin by reviewing these theories. Then we discuss the experimental evidence, particularly the recent literature on deception. II.1. Theory of Cheap Talk Game Information is often delivered in a strategic way. When the information holders do not have the same incentive as an uninformed decision maker, they tend to hold back some but not all of the information in order to gain an advantage in the transaction. This important economic intuition was first described in the seminal model by Crawford and Sobel (1982). In their paper, a sender has full knowledge of the state of the world and can send messages to influence a receiver’s belief so that he or she may make a choice that benefits the sender. The receiver, who is fully aware of the possibility of manipulation in senders’ messages, chooses an action that maximizes his or her own earnings. !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 4 !A!few!studies!look!at!environments!with!one!sender!and!two!receivers!(Battaglini!and!Makarov!,2011)!or!where! there!are!two!senders!and!one!receiver!(Minozzi!and!Woon,!2011;!Lai,!Lim,!and!Wang,!2011).!Those!studies!differ! from!ours,!as!players!in!those!experiments!make!decisions!as!either!a!sender!or!a!receiver,!but!never!both.!We! focus!on!a!game!that!better!describes!the!environment!of!intraXorganizational!communication,!which!is! characterized!by!having!each!player!act!as!both!sender!and!receiver.! ! ! The model implies, in equilibrium, larger the payoff differences between players lead senders to hold back more truthful information. In the limit senders are predicted to send random messages (engage in cheap talk). The seminal work by Crawford and Sobel (1982) has been extended in many directions. For example, Milgrom and Roberts (1986), Gilligan and Krehbiel (1989), Austen-Smith (1993), Krishna and Morgan (2001a, b) investigate the case where there is more than one sender for each receiver. Battaglini (2002) and Ambrus and Takahashi (2008) further extend the analysis to environments in which senders give advice on multidimensional issues. Morgan and Stocken (2008) study the case of polling in which each sender has a different information and ideology. Additionally, Farrell and Gibbons (1989) discuss the case where there are two receivers and two states of the world. In all of these cases, however, each agent plays either as receiver or as sender. Despite the important insights these models convey, they contrasts with the situations we are interested in on one very important dimension: players in these models are either a sender or a receiver, never both. The types of real world environments we are interested in would seem to contradict the model where the roles of a sender and a receiver are separated. For example, workers from different departments often talk and listen to each other, and people of one political party often express their own opinion to others and receive opinions from the other party. We are aware of only two models of strategic information transmission in networks where each person can act as both sender and receiver (Hagenbach and Koessler, 2010; Galeotti et al 2013). Many real world environments would seem to require this framework. For example, workers from different departments at the same company often talk and listen to each other, and people of different political opinions may also mutually exchange information. Hagenbach and Koessler (2010) investigated a case where each individual receives some information and the aggregation of all private signals reveals the truth. In their environment, a player earns more when (1) choosing a number that is closer to the true state of the world plus an individual bias5; and (2) choosing a number that is closer to others’ choices. The first part incentivizes the individuals to make the best guess of the truth, while the second half requires coordination of choices between players. As in other cheap talk games, players send messages free-of-cost before choosing numbers. Additionally, all messages are non-verifiable. Galeotti et al (2013) also models group communication; however, the earnings in their model are defined in a different way. In particular, their model assumes that a player earns the highest payoff if everyone in the game, including him/her self, chooses a number that matches the truth plus his/her own bias. Given that different players have different biases, one may try to affect others’ beliefs using cheap talk messages. The predictions of this and Hagenbach and Koessler (2010) are quite similar. Since we build primarily from Galeotti et al (2013), we make clear the details of their model in Section III. !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 5 !Just!as!in!Galeotti!et!al!(2013),!described!in!detail!below,!the!“bias”!in!Hagenbach!and!Koesslers!(2010)!is!a!payoffX relevant!parameter!that!can!differ!between!the!two!groups!and!that!can!affect!the!number!they!prefer!the!other! group!to!choose,!and!thus!can!impact!their!decision!regarding!whether!to!send!a!truthful!message.!! ! ! II.2. Deception Experiments The early experimental literature on cheap talk game studies the fit of empirical data to the predictions of Crawford and Sobel (1982). In those games, senders can choose vague messages by sending a range of possible states (e.g. sending (1-3) when signal is 2.). Dickhaut, McCabe and Mukherji (1995) confirms the comparative statics of the model by showing that the senders’ messages become vaguer and the receivers’ actions deviate more from the true state as preferences between sender and receiver diverge. Cai and Wang (2006) replicate the above finding and further show that the average payoffs of senders and receivers are very close to the predicted level for the most informative equilibrium. Their data also suggest that senders overcommunicate and receivers over-trust the message. Wang, Spezio and Camerer (2010) study the source of over-communication using eye-tracking data. Some recent experimental research uses a simplified sender-receiver game to study deception behavior in the lab. Gneezy (2005) analyze an experiment where there are only two states of the world. They find that people are sensitive to both their own gain and others’ losses when deciding to lie. Lundquist et al (2009) modify the game further into a labor contract context, where the senders have information on their ability level and face an incentive to lie so the receiver will agree to hire. With this design, they can observe not only whether a player has lied, but also the size of the lie. They find that lie aversion increases with the size of the lie and also the strength of the promise. The data also shows that free form messages lead to fewer lies and more efficient outcomes. Sheremeta and Shields (2013) designed a sender-receiver game where the subjects play the role of a sender or a receiver sequentially. With this design, the authors can identify whether the subject who lied as a sender will believe other’s message as a receiver. They find the liars believe and the lying behavior can be rationalized by accounting for elicited beliefs and other-regarding preferences. In this literature6, messages are typically considered deceptive if a sender’s message contains representations that differ from the true state of the world. We also use this method to analyze deceptive messages7. III. Theoretical Background Our experiment design is inspired by the two-group communication model in Galeotti et al (2013). We review the details of the model in this section. The set of players is denoted by N={1,2,…,n} partitioned into two groups, N1 and N2, with size n1 and n2, respectively, where n1+n2=n. Without loss of generality, assume n1 >n2≥1. Player i’s individual bias is bi. In the two-group communication model, each member of group 1 has a bias normalized to 0; members of group 2 have a bias bi=b>0. The state of the world ! is uniformly distributed on [0, 1]. Every player i receives a private signal si∈{0,1} where si=1 with probability!!. !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 6 !Sutter!(2009)!and!Xiao!(2012)!consider!deception!to!be!“sophisticated”!if!a!deceptive!sender!chooses!the!true! message!with!the!expectation!that!the!receiver!will!not!follow!his/her!message.! 7 !Identifying!a!player’s!exact!strategy!(truthXtelling!or!cheap!talk)!requires!repeated!observations.!The!result!may! be!ambiguous!if!individuals!switch!between!different!strategies!during!the!game.!We!do!not!try!to!identify!players’! strategies!but!simply!study!the!frequency!of!deceptive!messages.! ! ! Communication among players is exogenously restricted by a communication network ! ∈ {0,1}!×! where player i can send a message to j if gij=1 with gii=0 for all ! ∈ !. The communication neighborhood of i is the set of players to whom i can send his/her signals; it is denoted by Ni(g)={!! ∈ !: gij=1}. In this study we focus on the case where g is a complete network, meaning players can send a message to any other player. Communication mode describes to what extent the technology of communication allows for targeting messages. In a private message setting, player i chooses what message to send to each other player j. A communication strategy profile for each signal si∈{0,1} is defined as m={m1,m2,…,mn} in which mi(si)={m!" }!!∈!,!!! . After communication occurs, each player chooses an action. Agent i’s action strategy, based on his/her own signal and messages received from others, is yi:{0,1}n-1×{0,1}!R; y={y1,y2,…yn} denotes an action strategy profile. Given the state of the world ! and a profile of actions ! = {!! , !! , … , !! }, the payoff of i is: (!! − ! − !! )! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!(1) !! ! ! = − !∈! That is, agent i’s payoffs depend on how close his/her own action yi and the actions taken by other players are to his/her ideal action !! + !. A communication network g together with a strategy profile (m, y) induces a subgraph of g in which each link involves truthful communication. They refer to this network as the equilibrium truth-telling network denoted by c(m,y|g), a directed graph where !!" !, ! ! = 1 if and only if j belongs to i’s communication network and !!" ! = ! for every s={0,1}. Given c(m,y|g) and that the agents are divided into two groups, the in-degree of an arbitrary player in group i, !! is defined as the number of agents who send a truthful message to him/her. Among all the truthful messages, the number sent by members of the same group is denoted by kii, while the number sent by members of the opposite group is kij. Their analysis focuses on pure strategy Bayesian Nash equilibrium. They provide a full characterization of the utility-maximizing equilibrium networks8 with a focus on the natural subclass of those networks where there is complete intra-group communication. In our experiment setting, the bias we choose yields the same prediction whether we decide to use the full characterization or the subclass. The following equation describes the in-degree of an arbitrary player in group i in the utility-maximizing equilibrium truth-telling network: !!! = !! − 1;!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!(2) !!" = !"# !"# 1 − !! − 2 , !! , 0 , !, ! = 1,2, ! ≠ !;!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!(3) 2! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 8 !It!is!a!tradition!in!strategic!information!transmission!models!to!characterize!the!utility!maximizing!equilibrium!as! babbling!is!always!an!equilibrium!solution!but!not!meaningful!in!most!contexts.! ! ! That is, if ! < !> ! !(!! !!) ! !(!!!) , both intra-group and inter-group communication is complete; and if , there is complete intra-group communication and no inter-group communication. When b takes the intermediate value, inter-group communication also takes intermediate value9. Given this type of communication, in equilibrium all players trust all intra-group communication. ! They treat inter-group messages as true signals whenever b < , and as no information !(!!!) whenever!! > ! !(!! !!) . That completes the equilibrium prediction of the model. IV. Design and Procedure IV.1 Experiment Design The design of our experiment is based on the theory of Galeotti et al (2013) detailed above. Each experimental session includes 15 subjects. They are randomly assigned into three groups to play the game. Hence, there are 5 subjects in each group. The group plays the game repeatedly for a random number of rounds10 within a stage game. Players know that the other four players are fixed during the repeated game, and each of them holds a unique identifier: J, K, L, M or N. Players J, K and L belong to Group 1. Players M and N belong to Group 2. Group 1 and Group 2 players differ in their payoff function by only one parameter: the bias. The bias level also stay fixed for all rounds in a stage. Moreover, all the above information are common knowledge for all players. Once a stage game ends, 15 subjects are rematched into three new groups, assigned a new ID, given a different set of bias levels and restart a new stage. In total, all subjects experience three stage games by the end of the experiment. Each round of the experiment is a guessing game. Before a round starts, the computer generates a random integer r between 0 and 5 (including 0 and 5). The number is unknown to all players. At the beginning of each round, each player receives a private signal that is either 0 or 1. Players do not see others’ signals. However, they are told that the sum of the five signals received by all five players equals the random integer11. Before players guess the number, they are given the opportunity to exchange “cheap-talk messages” between each other. The messages are constrained to be either 0 or 1 to match the space of the signal. Moreover, messages are group-specific, so each player decides on what message to send to Group 1 and Group 2 players rather than sending a unique message to each player12. After all players submit their messages, !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 9 !Specifics!related!to!intermediate!biases!can!be!found!in!Galeotti!et!al!(2011).!! ! 10 !There!are!always!at!least!4!rounds!in!a!stage.!After!round!4,!the!game!has!a!random!stopping!probability!of!0.04! at!any!given!round.!To!keep!control!over!the!length!of!the!real!experiment,!we!randomly!generate!predetermined! round!lengths!of!19,!28!and!32!for!experimental!stages!I,!II!and!III!respectively.!The!practice!stage!lasts!3!rounds.!!! 11 !This!part!of!design!follows!Hagenbach!and!Koessler!(2010).!We!deviate!from!Galeotti!et!al!(2013)!for!two! reasons:!(1)!the!former!involves!less!uncertainty!and!thus!is!an!easier!task!for!our!subjects;!and!(2)!the!main! predictions!that!we!test!in!this!paper!remain!the!same!between!the!two!models.! 12 !The!message!is!group!specific!in!order!to!simplify!the!decision!problem!for!the!subjects.!! ! ! they observe the messages that are sent to them and are asked to guess the value r randomly chosen at the beginning of the round. They also choose a number x based on their guess of r to determine everyone’s payoff for that round. The payoff functions for Group 1 and Group 2 players are as follows13: 5 Payoff J , K , L = 20 − ∑ ( xi − r − b1 ) 2 (4) Payoff M , N = 20 − ∑ ( xi − r − b2 ) 2 (5) i =1 5 i =1 Players J, K and L share the same payoff function, as shown in equation 4. The payoff is maximized when all five players choose the number x that equals the true value of the random number r plus a group-specific bias b1. Players M and N share the same payoff function, as shown in equation 5. The difference between their payoff functions and those for Group 1 players is the group-specific bias b2. As indicated in the theory, this payoff structure incentivizes every player to: (1) choose a number x that is as close as possible to their best guess of r plus their own group’s bias b; and (2) make other players, both in the same group and in the other group, choose the same x. The presence of cheap talk messaging makes it possible for players in one group to manipulate the choice of x made by players in the other group. In our experiment setting, b1 and b2 can only take four different values, that is (0, 0), (0, 1), (1, 0) or (1, 1). Note that (1, 1) appears always and only in the practice stage, and thus is not included in our data analysis. The other three combinations appear in random order for the three experimental stages. The structure of the game and all payoff-related information, including the value of b1 and b2, are common knowledge. Players also know that the value of b1 and b2 remain fixed within a stage game, but change between stages. The following three screens implement this design. First, subjects send messages using the “messaging screen” (see Appendix A, Fig. 1). Then, subjects make guesses (that do not affect their payoffs) about the state r.14 Next, subjects choose the payoff relevant value of x using the “guessing screen” (See Appendix A, Fig 2). While they are making these two choices, the same screen also shows them the messages they received from others graphically. Finally, the “result screen” (see Appendix A, Fig 3) reveals the true value of the random integer and displays all the actions taken by the other four players and their current payoff. Payoffs accumulate within, but not between, each of the three stage games. Players are informed about their accumulated payoff at the end of each stage. They are also reminded that they will be re-matched with a new set of players, and that their stage payoff will not be carried over to the new stage. Each subject’s earnings for the experiment are determined by one randomly-determined stage game according to a die roll at the end of the experiment. !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 13 !The!payoff!differs!from!the!theory!section,!as!we!give!20!experimental!dollars!as!an!endowment!per!period.!This! ensures!subjects!do!not!earn!negative!amounts!during!the!experiment.!This!change!does!not!alter!the!theoretical! predictions.!! 14 !The!first!guess!does!not!affect!payoffs!but!is!used!as!a!way!to!verify!that!participants!take!account!of!the!bias! when!making!their!payoff!relevant!choice.!In!particular,!the!payoff!relevant!choice!should!be!equal!to!their!guess! plus!the!bias.!Nearly!all!of!our!subjects!displayed!the!expected!relationship!between!their!guess!and!their!payoffX relevant!choice.! ! ! Note that our experiment does not strictly follow Galeotti et al (2013). One difference is that Galeotti et al (2013) assumes that the true state is uniformly distributed, and that players signals are drawn iid from {0,1} with probability that the signal is “1” equal to the value of the state. In our experiment the true state is also drawn from a uniform distribution on the discrete state {0, 1, 2, 3, 4, 5}, and each player receives a signal that, ex-ante, is equal to “1” with probability equal to one-fifth the value of the state. Consequently, in an ex-ante sense, all players understand they are equally likely to receive a signal of “1”, and that the likelihood varies depending on the true state. It is easy to verify that the theorems in Galeotti et al (2013) require only that this be true. The fact that there is correlation in our signals among players does not impact equilibrium predictions. The intuition is that play is simultaneous, and the state is unknown, meaning there is no way for players to exploit information about correlation in their actions, either in theory or in practice. A second difference is that we restricted message sending so that each player was required to send the same message to all members of the a given group. Note that the impact of this is to eliminate some types of off-equilibrium play. While in some games restricting offequilibrium play can change equilibrium outcomes (e.g., punishment games), that is not the case here. It is easy to verify that restricting off-equilibrium decisions does not in this case change the game’s unique equilibrium. IV.2 Procedures The experiment sessions were conducted between May 2012 and June 2012 in the ICES laboratory at George Mason University and in April, 2014 in the experimental lab at Weber State University15. Subjects were recruited via email from registered students at George Mason University and Weber State University. Each subject participated in only one session and none had previously participated in a similar experiment. The result from either institution does not differ from each other in any meaningful way, so we pool all the data in our analysis. In total, 90 subjects participated in the computerized experiment programmed with z-Tree (Fischbacher, 2007). Each experimental session lasted between 120 and 150 minutes. Subjects’ total earnings were determined by the Experimental Dollars (E$) earned at the end of the experiment, which were then converted at a rate of E$20 per US dollar. Average earnings were $25.27, ranging from a maximum of $27.5 to a minimum of $18 across all sessions (excluding the $5 show up fee). In all treatments, before a session began, subjects were seated in separate cubicles to ensure anonymity. They were informed of the rules of conduct and provided with detailed instructions. The instructions were read aloud. In order to ensure there was no confusion, after subjects finished reading the instructions, they were asked to complete a quiz. An experimenter checked their answers and corrected any mistakes one by one. Then the experimenter worked through the quiz questions on a white board in front of all subjects. The experiment began after !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 15 !In!response!to!a!revision!request,!we!ran!three!additional!sessions!at!the!experimental!lab!in!Weber!State! University,!where!one!of!the!authors!is!affiliated!with!at!the!time.! ! ! all subjects confirmed they had no further questions. All sessions in both institutions were conducted by the same experimenter to ensure the procedure in use is the same. 90 subjects participated in the experiment during six sessions. Within each session, we obtained 97 message-sending decisions from each subject (excluding the practice stage). Since we rematch groups between stages, our analysis conservatively assumes the average measure at the unique group level within each stage is an independent observation. Hence, our data analysis is based on 54 observations. V. Hypothesis and Results V.1 Hypothesis Based on the theory of Galeotti, et al (2013) discussed in section III, we make the following hypothesis: Hypothesis 1: Players always tell the truth to those in the same monetary group. Hypothesis 2: Players tell the truth to others in a different monetary group if the bias is (0, 0), and always send random message (babble) if the bias is (0,1) or (1,0). Hypothesis 3: Players fully trust the message sent by their group members and also fully trust the message sent by other group members if bias is (0, 0). They disregard those betweengroup messages if the bias is (0,1) or (1,0). V.2 Results We lay out the results in the order of the hypotheses above. First, we show the choice of message-sending (to test Hypothesis 1 and 2). Then, we investigate the behavior of guessing number (to test Hypothesis 3). Result 1: Most within-group messages are truthful. Our data support hypothesis T1. As shown in Figure 1, we find that 95.63% of the within-group messages are truthful. Although the overall level of truth-telling seems high, it is significantly lower than the predicted level of 100% (Wilcoxon signed-rank test, p<0.001). Consistent with the theory, the bias of the opposing group does not affect within-group messages in any statistically significant way (Mann-Whitney ranksum test, pairwise comparisons, all pvalues greater than 0.2516). Moreover, group size also does not impact the truthfulness of within-group messages (95.7% for Group 1 and 94.8% for Group 2, Mann-Whitney ranksum test, p=0.2130). Result 2: Players tell less truth to members of different group than to their own group members. Our data support hypothesis T2. 83.4% of messages sent between two groups are truthful. This level of truth-telling is much lower in comparison to within-group messages (Wilcoxon sign-rank test, p<0.001). This effect is larger if the bias is (0,1) or (1,0). However the effect persists even if the bias is (0,0). ! ! In the case where bias is (0, 0), the two groups share the same payoff function, so that truth-telling is predicted to be 100%. However, 90.68% of these between-group messages are truthful, a significantly lower percentage than predicted (Wilcoxon signed-rank test, p<0.001). It is also lower than the truthfulness for within-group messages (compare to 95.63%, Wilcoxon sign-rank test, p<0.03), suggesting that simply dividing subjects into two groups has an impact on their truthfulness regardless of monetary incentive16. According to theory, under bias (0,1) and (1,0), there only exists a babbling equilibrium with 50% truthful messages. We observe 79.77% truthful messages between groups17, which is significantly higher than predicted levels (Wilcoxon signed-rank test, p<0.001). Under either bias, truth-telling is significantly lower than the case where the bias is (0,0) (pairwise comparisons, Mann-Whitney ranksum test, p=0.0001 and 0.0075 respectively). Result 3: Players overly trust messages they receive. To measure whether a player believes the messages s/he receives, we measure the difference between one’s guess and the sum of “1” messages received. When the difference is zero, we define the “trust” measure to equal one and set it to zero otherwise. 80.59% of all guesses submitted exactly equaled the sum of “1” messages received. When the bias is (0,0), !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 16 !Eckel!and!Grossman!(2005),!however,!suggest!that!minimal!group!identity!does!not!affect!subjects’!behavior!in! their!experimental!setting.!Our!data!suggests!that!the!effectiveness!may!be!sensitive!to!the!environment.! 17 !We!combine!the!two!cases!together!as!the!unequal!bias!case,!as!there!is!no!significant!difference!between!them! (p=0.652)! ! ! 85.58% of guesses are consistent with the messages received, which is significantly lower than the predicted 100% level of trust (ttest, p<0.001). In equilibrium, when the bias is either (0,1) or (1,0), random choice will lead Group 1 players to appear completely trusting of between-group messages 37.5% of the time. This can be seen as follows. In equilibrium, each player in Group 1 faces four possible message combinations sent at random by two Group 2 players: (0,0), (0,1), (1,0) and (1,1). Each of these four outcomes is equally likely to appear. Group 1 players form beliefs about the true signal that Group 2 players hold independent of these messages: (0,0), (0,1), (1,0) and (1,1). Each outcome is also equally likely to happen. Therefore, out of 16 message-belief pairs with each combination having the same probability, six of the sums can coincide at random (6/16=37.5%). Similarly, Group 2 players may appear to be trusting 31.25% of the time even if they are choosing at random. Overall then, random choice will lead 35% of choices to appear fully trusting. Our data show that 76.63% of guesses are fully trusting, significantly higher than 35% (ttest, p<0.001). Moreover, the trust levels between bias (0,0) and bias (0,1) and (1,0) are significantly different (ttest, p<0.001). VI. Conclusion In a great number of socio-economic environments, information is transmitted between group members strategically. This may occur in part due to monetary incentives, which can affect strategic considerations, and ultimately the truthfulness of people’s messages. Based on a model suggested by Galeotti et al (2013), we conducted a laboratory study of deceptive behavior in an environment of strategic information transmission. In our design, two groups of subjects with different monetary incentives could attempt to influence each others’ decisions by sending potentially dishonest messages. We found that the message-sending behavior of our subjects conformed to theory. In particular, within-group messages were mostly truthful and betweengroup messages were relatively less truthful. We found behavior to depart from predictions, however, in that we find more truth-telling between groups with misaligned incentives than theory predicts. The decision to behave honestly when deception is economically optimal is an often reported finding in the behavioral economics literature. An explanation is that people have social image concerns that mitigate their dishonesty (Hao and Houser, 2013). Whether and how social image might play a role in explaining data in these environments is a topic worthy of additional exploration. The results of this study suggest that misaligned incentive may explain the intraorganizational conflict we often see in business practice. To reduce the conflict and deceptive communication, it may be of benefit to develop methods to reduce incentive differences between departments. As monetary incentives may be costly or sometimes impossible to manipulate, studying social incentives and behavioral anomalies, as well as the role of emotion and social preferences in these environments, might be a profitable directions for future research. ! ! Reference: Ambrus, A. and S. Takahashi (2008): Multi-sender Cheap Talk with Restricted State Spaces, Theoretical Economics, 3:1-27 Austen-Smith, D. (1993): Interested Experts and Policy Advice: Multiple Referrals under Open Rule, Games and Economic Behavior, 5(1): 3-43 Battaglini, M. (2002): Multiple Referrals and Multidimensional Cheap Talk, Econometrica, 70(4): 1379–1401 Battaglini, M., and U. Makarov (2011): Cheap Talk with Multiple Audiences: an Experimental Analysis, working paper Blume, A., D. DeJong, Y. G. Kim and G. Sprinkle (1998): Experimental Evidence on the Evolution of Meaning of Messages in Sender-Receiver Games, American Economic Review, 88: 1323-1340 Blume, A., D. DeJong, Y. G. Kim and G. Sprinkle (2001): Evolution of Communication with Partial Common Interest, Games and Economic Behavior, 37: 79-120 Cai, H and J.T. Wang (2006): Overcommunication in Strategic Information Transmission Games, Games and Economic Behavior, 56 (1):7-36 Cloke, K., J. GoldSmith (2000): Resolving Personal and Organizational Conflict: Stories of Transformation and Forgiveness, Jossey-Bass Conrad, Charles R., M.S. Poole (2011): Strategic Organizational Communication: In a Global Economy, Wiley & Sons Cowan, David. (2003): Taking Charge of Organizational Conflict: A Guide to Managing Anger and Confrontation, Personhood Press Crawford, V.P. and J. Sobel (1982): Strategic Information Transmission, Econometrica, 50(6):14311451 Dana, J., R. A. Weber and J. X. Kuang (2007): Exploiting moral wiggle room: experiments demonstrating an illusory preference for fairness, Economic Theory, 33: 67–80 De Dreu, C.K.W., M. J. Gelfand (2007):The Psychology of Conflict and Conflict Management in Organizations, Psychology Press Dickhaut, J.W, K.A. McCabe and A. Mukherji (1995): An Experimental Study of Strategic Information Transmission: Economic Theory, 6:389-403 Eckman, A., T. Lindlof (2003): Negotiating the Gray Lines: an ethnographic case study of organizational conflict between advertorials and news, Journalism Studies, 4:65–77 Farrell, J. and R. Gibbons (1989): Cheap Talk with Two Audiences, American Economic Review, 79(5): 1214-23 Galeotti, A, C. Ghiglino and F. Squantani (2013): Strategic Information Transmission in Networks, Journal of Economic Theory, 148(5): 1751-1769. ! ! Gilligan, T. and K. Krehbiel (1989): Asymmetric Information and Legislative Rules with a Heterogeneous Committee, American Journal of Political Science, 459-90 Gneezy, U. (2005): Deception: The Role of Consequences, The American Economic Review, 95(1): 384-394. Gupta, A. K., S. P. Raj, and D. L. Wilemon (1985): R&D and Marketing Dialogue in High-Tech Firms. Industrial Marketing Management 14, 289–300 Hagenbach, J. and F. Koessler (2010): Strategic Communication Networks, Review of Economic Studies, 77(3):1072-1099 Hao, L. and D. Houser (2013) “Perceptions, Intentions, and Cheating”, working paper Kolb, D. M., L.L. Putnam, J. M. Bartunek (1992): Hidden Conflict in Organizations: 1st Edition, SAGE Publications Krishna, V. and J. Morgan (2001a): A Model of Expertise, Quarterly Journal of Economics, 116(2):747-775 Krishna, V. and J. Morgan (2001b): Asymmetric Information and Legislative Rules: Some Amendments, American Political Science Review, 95(2):435-452 Lai, E. K., W. Lim, and J. T.-Y. Wang (2011): Experimental Implementations and Robustness of Fully Revealing Equilibria in Multidimensional Cheap Talk, working paper Lundquist, T., T. Ellingsen, E. Gribbe and M. Johannesson (2009): The Aversion to Lying, Journal of. Economic Behavior and Organization, 70(1–2):81–92 Milgrom, P.R. and J. Roberts (1986): Relying on the Information of Interested Parties, Rand Journal of Economics, 17: 18-32 Miller, Katherine (2011): Organizational Communication: Approaches and Processes, Cengage Learning Minozzi, W., and J. Woon (2011): Competition, Preference Uncertainty, and Jamming: A Strategic Communication Experiment, working paper Morgan, J. and P. C. Stocken (2008): Information Aggregation in Polls, American Economic Review, 98(3): 864-96. Pirnejad, H., Z, Niazkhani, M.Berg and R. Bal (2008): Intra-organizational Communication in Healthcare--Considerations for Standardization and ICT Application, Methods of Information in Medicine, 47(4): 336-45 Rahim, Afzalur (2000): Managing Conflict in Organizations: 3rd Edition, ABC-Clio, LLC Sheremeta, R. and T. Shields (2013):
Do Liars Believe? Beliefs and Other-Regarding Preferences in Sender-Receiver Games,” Journal of Economic Behavior and Organization, 2013, 94, 268-277. Sutter, M.(2009): Deception through Telling the Truth? Experimental Evidence from Individuals and Teams. Journal of Econonomics. 119, 4760. ! ! Tobak, S. (2008): Marketing v. Sales: How To Solve Organizational Conflict, CBS, MONEYWATCH Wang, J.T., M. Spezio and C.F. Camerer (2010): Pinocchio's Pupil: Using Eyetracking and Pupil Dilation to Understand Truth Telling and Deception in Sender-Receiver Games, American Economic Review, 100 (3): 984-1007 Weinrauch, J. D., R. Anderson (1982): Conflicts Between Engineering and Marketing Units, Industrial Marketing Management, 11(4): 291-301 Xiao, E. (forthcoming): "Profit Seeking Punishment Corrupts Norm Obedience, Games and Economic Behavior ! ! Appendix A. Z-tree Interface Figure 1. Messaging Screen ! ! Figure 2. Guessing Screen ! ! Figure 3. Result Screen ! !