Dattner Consulting

Papers: Trust and Negotiation

Introduction

The "mythical fixed pie" is a well documented phenomenon- people believe that negotiation situations are zero-sum, even when there is potential for both parties to create value, not just claim value. Theorists believe that the mythical fixed pie is the result of an overextension of the societal value of competition and the affect associated with that value, and that it serves as a heuristic to simplify cognitively complex negotiation situations. (Pruitt & Rubin, 1986; Bazerman and Neale, 1983, Bazerman, Magliozzi & Neale, 1985). The presence of the mythical fixed pie can be gauged by measuring fixed sum error, the extent to which negotiators fail to perceive potential for integration. (Pinkley, Griffith and Northcraft, 1995).

Previous research has demonstrated that positive affect can counteract the mythical fixed pie (Carnevale & Isen, 1986), but no previous research has examined whether and under which circumstances negative affect can make perceptions of a mythical fixed pie more likely to occur. The present study will examine whether and to what extent anxiety and self-protectiveness cause negotiators to incorrectly process information and to overlook the potential for the integration of their interests.

In the present study, anxiety will be operationalized as gain versus loss frame - the negotiator's perception of the situation as one in which there is the possibility of either a positive or a negative outcome, and self-protectiveness will be operationalized as trust versus mistrust. The loss frame can be thought of as inducing negative affect about the negotiation situation, which is the background upon which mistrust can be thought of as inducing negative affect about the other negotiator.

It is hypothesized that because the underlying mechanisms of anxiety and self-protectiveness limit cognitive flexibility and lead to reliance on zero-sum cognitive heuristics, there will be main effects for both independent variables, with loss frame (anxiety) leading to higher fixed sum error than gain frame and mistrust (self-protectiveness) leading to higher fixed sum error than trust.

It is also hypothesized that due to the fact that in a negotiation, gain or loss frame provides a situational context in which the other negotiator is perceived, there will be an interaction between trust and frame wherein trust will decrease fixed sum error more in a loss frame than in a gain frame. In a low anxiety gain frame, negotiators are likely to feel positively about the negotiation and the potential outcome, and therefore to be more flexible in their perceptions of the possibility for integrative agreement, regardless of their level of trust for the other party. However, negotiators in a loss frame are likely to be wary, and therefore to care much more about the trustworthiness of their counterparty. In a situation where the loss-framing negotiator trusts his or her counterpart, he or she will be more relaxed, flexible, and open to perceiving the potential for integrative agreement than if trust were absent. In other words, it is hypothesized that the affect induced by the situation moderates the affect induced by the other negotiator, and that their main effects and interaction all influence information processing.

Pinkley, Griffith and Northcraft (1995) presented two possible explanations for fixed sum error: information availability errors and information processing errors. Information availability errors occur when information that could suggest integrative agreement is not presented by the negotiating parties. Information processing error is when information that could suggest the possibility for integrative agreement is presented, but is ignored or distorted.

The present study will focus on the role of trust in information processing and will test the hypothesis that anxiety and self-protectiveness lead to faulty processing of available information. All the information sufficient for the negotiators to realize that there is the potential for integration will be made available, thereby isolating any effects of the independent variables on the processing of that information.

Previous research has provided evidence that there can be an interaction between stress and trust in which trust only matters in the higher-stress situation. However, that research has been limited to studies of information availability. For example, Pruitt (1981) conducted a 2 X 2 experiment in which the independent variables were competition, as operationalized by high versus low limits, and high trust versus low trust. The dependent variable was the amount of information the negotiators shared with one another. The author found that trust only made a difference in the competitive high limit case. In the non-competitive low limit case, there was no difference between high and low trust.

The findings of Parks and Hulbert, (1995), also suggest that trust is more important in a stressful situation than in a non-stressful one. They conducted a 2 x 2 experiment which investigated the impact of trust on negotiation process. They investigated whether high versus low trusters, as dispositionally defined, would cooperate equally under conditions of fear (of receiving no payoff) versus no fear. They found that "differences between high and low trusters was observed when fear was present, but that no differences existed when fear was absent." (p.725).

The present study will determine if the findings of Pruitt (1981) and Parks and Hulbert (1995) - that the importance of the perception of the other negotiator (mistrust versus trust) is moderated by the perception of the negotiation situation (stress/fear/loss frame versus low stress/no fear/gain frame) - will also hold true with a new dependent variable, fixed sum error, which will be caused by information processing errors on the part of the negotiators, rather than by information availability constraints.

There are many competing conceptions of trust. Lewicki and Bunker (1995) write: "given the level of interest in trust across the disciplines, one might be tempted to assume that a coherent and integrated literature exists. Unfortunately, nothing could be further from the truth." (p. 135) The authors claim that theories of trust are fragmented because they have been developed in different theoretical paradigms.

For the purposes of this study, the independent variable of trust will be operationalized using Rotter's (1967) classic conception of trust as "an expectancy held by an individual or a group that the word, promise, verbal or written statement of another individual or group can be relied upon." (p. 651). Although Rotter approached trust from a dispositional perspective, in the present study trust will be a state, not a trait. Gain versus loss frame will be operationalized as it was in "prospect theory", first published by Kahneman and Tversky (1979).

The dependent measure will be fixed sum error. In general, fixed sum error, a quantitative indicator of how much potential for integration negotiators fail to perceive, occurs because human negotiators diverge from optimal rationality, because they are unable, in the words of Bazerman and Carroll (1987) to identify "all alternatives and all outcomes, gather all relevant information, and combine it optimally" the reason being that "negotiators behave in a more selective, abbreviated and even biased manner". (P. 252)

In the present study, subjects will be presented with a simulated negotiation situation between the management and the union of a company which will contain the potential for integrative agreement, both within issues and between issues. The independent variable of trust vs. mistrust will be manipulated by providing subjects with a "record" of whether their "opponent" honored or did not honor agreements made in previous negotiations. The gain frame vs. loss frame will be manipulated by an instruction sheet that will emphasize outcomes as either losses or gains. The dependent variable will be fixed sum error, in this case a point value that reflects the extent to which a subject has incorrectly processed information that indicates potential for integration in the negotiation.

Method

Subjects and Design

Subjects were 98 undergraduates at New York University, enrolled in introductory psychology, and they received partial fulfillment of a course requirement for their participation.

The experiment was a 2 (gain versus loss frame) X 2 (trust versus mistrust) between subjects design. The experiment involved a simulated negotiation between the management and the union of a car company. There were four issues that the subjects were negotiating about- wages, medical benefits, vacation time, and job training. Subjects were randomly assigned to one of four conditions- gain/trust, gain/mistrust, loss/trust, loss/mistrust as well as to union or management.

Procedure

Subjects signed up for half hour time slots, each of which was open to two subjects. Upon arrival in the laboratory they were seated facing one another at a table and they each filled out a consent form. They were then given both verbal and written instructions. The verbal instructions, given by the experimenter, were a brief orientation to the experiment, in which one subject was informed that he or she would represent the union while the other subject was informed that he or she would represent management, and a description of the written material in front of the subjects as well as the procedure of the experiment. Subjects were instructed both verbally and in the written materials to not speak to one another until they had filled out a pre-negotiation questionnaire.

After the verbal instructions, subjects read two pieces of paper that were the manipulations of gain versus loss frame and trust versus mistrust respectively. Subjects then read the point matrix for union and management and answered a pre-negotiation questionnaire. After they had both completed the questionnaire, they negotiated with one another, and then they completed a post-negotiation questionnaire which contained manipulation checks.

The negotiation

In the point matrix that was provided to the subjects there were three levels for each issue represented in rows, for example, wages could be $14 per hour, $15 per hour or $16 per hour. Subjects were given three point values for each level of each issue in columns, the points for themselves, for the other negotiator, and the total of their points and the other negotiator's points. The relationship between the point values for each negotiator was symmetrical and as follows: wages were diametrically opposed, meaning that one party's gain was the other party's loss; job training was congruent, meaning that both negotiators stood to gain from more job training; and medical benefits was more important to the union and vacation time was more important to management. Subjects were told that there would be a $50 raffle at the end of the semester in which their odds of winning would be determined by the number of points that they received after the negotiation.

Subjects were given written instructions that informed them that after the negotiation they would have the opportunity to either honor the agreement they reached or not honor it. There was a payoff matrix that indicated the different scenarios associated with whether or not they and the other negotiator honored the agreement. This was analogous to a prisoner's dilemma, in which the point values indicated that if a subject believed the other negotiator would honor, then his or her best strategy would be to honor, but not honoring would be the best strategy if the negotiator believed the other negotiator would not honor.

The independent variable manipulations were as follows: for loss versus gain frame subjects were given a point matrix that presented the post-negotiation honor or not honor scenarios as containing either the potential for gain of raffle points or for loss of raffle points. Trust versus mistrust was manipulated by presenting a false record of the other negotiator's prior negotiation record. In the trust condition, the record stated that the other negotiator had participated twice before and had honored both agreements. In the mistrust condition, the record stated that the other negotiator had participated twice before and had not honored both times.

The dependent variable was the total point value that the negotiators picked as the maximum joint benefit to both themselves and the other negotiator for all four issues. The correct solution involved splitting wages evenly (taking the middle level of this issue in a distributive way), the highest level of job training, and a "log-roll" trade-off between medical benefits and vacation time in which the union got its preference in medical benefits and management got its preference in vacation time. All information necessary and sufficient for subjects to find this solution was provided in the "total points" column.

Results

Manipulation checks

Manipulation checks were inconclusive with respect to the independent variables. In order to avoid the demand that would have been introduced by asking outright "did you trust the other negotiator?" the question was phrased: "If the other negotiator has participated in negotiation experiments before, did s/he honor the agreement(s)? Yes or No". Since the manipulation checks were collected after the negotiation, there is no way of knowing whether subjects verified with one another that neither in fact had participated before.

Likewise, the check for loss versus gain frame was "In this experiment, there was 1. the possibility of gaining raffle points or 2. the possibility of losing some of the raffle points that were assigned to you after the negotiation.". This was problematic in that the answer to this question is contingent on what happened during the negotiation. The possibility of gaining or not gaining points is contingent on whether an agreement was reached and whether one or both negotiators honored the agreement. In other words, the intervening negotiation rendered it very difficult to asses the subjects' state of mind when it mattered- before the negotiation when the dependent measures were collected.

Manipulation checks were also incompletely filled out by 33 out of the 98 subjects. Manipulation checks for the two independent variables were filled in "correctly" by 63% of the subjects in the gain/trust condition, 56% of the subjects in the gain/mistrust condition, 13% of the subjects in the loss/trust condition, and 8% of the subjects in the loss/mistrust condition.

Statistical Analyses

The correct answer to the question in which the dependent variable was collected was 2300 points- the maximum joint gain that was possible if integrative potential was perceived. The absolute value of the difference between the number provided by the subject and 2300 is the dependent variable- fixed sum error.

No significant results were obtained in an analysis of variance with frame and trust as independent variables and fixed sum error as the dependent variable. F (1, 94)for the main effect of trust was .615, p <.435, F (1, 94) for the main effect of frame was .165, p .686, and F (1, 94) for the interaction was .584, p < .447.

In order to determine whether the independent variables had an impact on the likelihood that subjects would discover that job training was a congruent issue, an analysis of variance was conducted in which frame and trust were the independent variables and points for training was the dependent variable. F (1, 94)for the main effect of trust was .702, p <.404, F (1, 94) for the main effect of frame was .004, p < .951, and F (1, 94) for the interaction was .342, p < .560.

An additional analysis of variance was run in which manipulation checks rather than conditions were used as independent variables. F (1, 62)for the main effect of "the other subject honored (or did not honor) in the past" was .093, p <.762, F (1, 62) for the main effect of "the negotiation contained the potential to gain tickets (or to lose) some of the tickets that you had received after the negotiation" was .006, p < .938, and F (1, 62) for the interaction was .131, p < .718.

Discussion

There are several possible explanations for why the hypotheses did not receive support. The subjects, undergraduates who were unfamiliar with negotiation situations, may have found the amount of information that they were provided with to be overwhelming. Differential comprehension by subjects might have introduced random error that might have obscured the effects of the manipulations. Subjects may not have understood the relationship between levels of issues and their point values versus the other negotiator's point values.

It is also possible that the manipulation of the independent variables were not effective due to a lack of credibility. In some cases, subjects may have known that the other had not participated in previous experiments, which would have nullified the intended introduction of trust or mistrust. 16 of the 98 subjects knew each other and/or were friends before the experiment which also may have influenced their perceptions of the situation and of one another.

Subjects also may not have believed that there was going to be a $50 raffle at the end of the semester, which may have obscured the impactfulness of the gain versus loss manipulation. Further, regardless of the presentation of the "gain" or "loss" outcome matrix, if subjects did believe that there would be a $50 raffle. subjects might have all been in gain frames since they might not have believed that the experiment contained any real possibility of losing something.

Out of the 98 subjects, only one did not honor the agreement. This suggests that social desirability or other unknown factors may have played a more significant role in subjects' decision making than did the independent variables. It is possible that these other factors also affect or obscure the relationship between the independent variables and the dependent variable.

Since the manipulation checks were problematic, no definitive statements can be made about why no significant results were obtained.

Conclusion

The findings of Parks and Hulbert (1995) and Pruitt (1981) suggest that stress moderates trust such that trust matters more in a high stress situation. The present study failed to replicate their findings, or to extend them to information processing rather than information availability.

A replication of the present study with a different population of subjects, such as MBA or law students, might eliminate the random error caused by incomprehension and allow for an isolation of the effects of the independent variables. If undergraduates were to be used again, the present study could be re-designed in order to be simpler and easier to comprehend, which could help reduce random error and isolate the impact of the independent variables on the mythical fixed pie.