Negotiator Consistency,Counterpart Consistency,and Reciprocity in Behavior Across Partners: A Round-Robin Study

Actor Effects: Negotiator Consistency

In the SRM, actor effects represent behavioral consistency. Consistent with interactionist theories of personality, we predict that individuals maintain at least some degree of unique behavioral signatures across their interactions with various partners. For example, negotiators who stand firm and indicate that they are willing to walk away may be generally likely to do this across their multiple partners. This is not to say that their behavior is identical across encounters, but rather that there is a degree of consistency that shines through. Outside of the negotiations domain, Leikas et al. (2012) had confederates enact a wide range of behaviors and found that individuals who interacted with these confederates were still rank-order consistent in spite of their partners’ differing interaction style.

Partner Effects: Partner Consistency

Although personality is partly defined by what an individual regularly does, personality is also characterized by what an individual regularly elicits other people to do (Buss, 1987; Kenny, 1994; Mischel, 2004). Intuitively, many common adjectives for describing people are less about the people themselves than about the effect the people have on others, for example, boring, annoying, and amusing. In the realm of emotion, Eisenkraft and Elfenbein (2010) found evidence for trait affective presence, or the tendency of individuals to elicit consistent emotions in their interaction partners, even apart from the emotions they themselves tended internally to feel. In the negotiations domain, we hypothesize that individuals’ counterparts may behave consistently as well. In the SRM, these tendencies are called partner effects. For example, some negotiators might tend to have partners who stand firm against them, regardless of their own tendency to stand firm.

At the same time that we hypothesize the existence of behavioral consistency among the various counterparts of a given focal negotiator, we acknowledge that past evidence is relatively weak for partner effects in behavior more generally. Kenny and Malloy’s (1988) review concluded that partner effects are small and sometimes elusive. They speculated that such effects are difficult to uncover due to the typical methodology in which strangers engage in unstructured social context outside of a context in which the measured variables would be expected to emerge. We maintain that the negotiations context may hold more promise for finding partner effects than past studies. In the present work, participants are goal-directed and focused on accomplishing a specific task that is engaging and has substantially greater length, and the measured behaviors directly relate to this negotiation task. As such, we are optimistic to find rare evidence for partner effects.

Dyadic Reciprocity: Behavioral Similarity

As mentioned above, one reason to expect systematic departures from behavioral consistency is the tendency of negotiators to reciprocate and complement their partners’ tactics (Weingart et al., 1990). As such, in addition to focusing on consistency, we also examine the dynamic behavioral interplay between the two parties. In particular, we hypothesize that negotiators tend to exhibit behavioral similarity rather than dissimilarity (i.e., complementarity) and rather than a lack of association. That is, parties reciprocate the behavior of the other. In the SRM, this is called dyadic reciprocity. Past work provides examples of negotiation partners imitating each other’s behavior (Axelrod, 1984; Deutsch, 1973; Lawler & Yoon, 1996; Putnam & Jones, 1982; Weingart et al., 1990). For example, they tend to reciprocate seeking and providing information (Thompson & Hastie, 1990; Weingart et al., 1993), and even reciprocate conflict situations that cause unproductive negative spirals when nonreciprocity would be more productive to fix the situation (Brett et al., 1998). Our current study offers two extensions to this past work. First, we expand greatly the range of behaviors examined for potential reciprocity. Second, we examine behaviors that are related to cooperation and competition in the context of a new transaction versus existing dispute. That is, the research on conflict spirals has taken place in the context of conflicts that already have a history of negative relationships, whereas this study examines deal making within a new relationship without problematic history.

Negotiation Behaviors

Consistent with Neale and Northcraft’s (1991) Behavioral Negotiation Theory, researchers have examined the tactics and behaviors that characterize more effective negotiation outcomes relative to less effective outcomes (e.g., Adair & Brett, 2005; Weingart et al., 1990). We include as many behaviors as possible that the existing literature has established as important for the negotiation context (Adair & Brett, 2005; Pruitt & Lewis, 1975; Weingart et al., 1990, 1993).

Relationship Quality

In addition to negotiation-specific behaviors, we examined negotiation partners’ nonverbal behavior. Given the value of relationship building, we measured the appearance of interpersonal rapport between the two counterparts, using Bernieri et al.’s (1996) validated technique. We also measured interpersonal synchrony, alternately called interpersonal coordination or behavioral entrainment. Synchrony has been implicated in many important phenomena of social interaction (Bernieri & Rosenthal, 1991), including negotiation (Curhan & Pentland, 2007). Synchrony occurs when an individual coordinates their rhythms and movements to those of an interaction partner and, together, the interaction creates a perception of physical unity and coordination (Bernieri et al., 1994; Condon & Ogston, 1966). Synchrony has multiple functions in the service of interpersonal affiliation. First, it helps to regulate interactions to keep them running smoothly (Bernieri et al., 1988). Second, it serves a communication function, signifying the interest or approval of an interaction partner (Kendon, 1970). Synchrony correlates with self-reported rapport between individuals (Bernieri et al., 1988; Tickle-Degnen & Rosenthal, 1990). In this study, we code for both synchrony and perceived rapport.

Language Use

How people use words is an established indicator of personal functioning. Based on a wide range of categories, word count serves as a proxy for content analysis and evidence demonstrates that it reveals information about an individual’s psychological functioning. This information includes attitudes, cognitive styles, relationships, and even physical and mental health (Pennebaker et al., 2003; Pennebaker & Graybeal, 2001; Pennebaker & King, 1999). The dictionary developed by Pennebaker and Graybeal (2001), called the Linguistic Inquiry Word Count (LIWC), has been extensively validated as a comprehensive tool. The LIWC system provides a “fingerprint” of an individual’s word usage and, in doing so, can serve as a marker of individual functioning (Pennebaker & Graybeal, 2001). Many of the items in LIWC are relevant to the negotiation setting. For example, the use of pronouns can serve as an indication of self-focus (I + self), other-focus (you), and relational-focus (we). Cognitive process words—such as the use of causal language—can indicate how negotiators think about their activities in terms of the framing of their deals. Verb tense (past, present, future) can also give a sense of negotiator focus. Communication style can give a sense of negotiators’ state of mind, for example, the extent to which they are tentative or certain, and based on their level of fluency in communication. Communication quantity can serve as an indicator of comfort and/or dominance in the interaction. Some words in the LIWC system relate to conversational flow, such as negations and assent. In this study, we make use of core components of the LIWC dictionary (Pennebaker & Francis, 1999) with a focus on concepts with potential relevance to negotiation, while excluding the list of specific topics such as time, family, and music. Given our overarching interest in comprehensiveness, our approach was to identify a long list of LIWC dictionary terms with prima facie relevance to negotiation rather than making specific predictions for every item on the list.

Participants

Data were collected from full-time MBA students at a large public university in the United States in the context of their elective course on negotiations. Procedures were monitored by the university’s Institutional Review Board (IRB), approved on the basis of pedagogical value and with participants completing documents of informed consent. A total of N = 103 people completed all measures below (72 male, 31 female; age M = 30.1 years, SD = 3.7), as one of their first classroom assignments. Although the SRM does not have formal power tests, we note that the design yielded 153 unique negotiations completed by the 103 participants, which is the equivalent statistical power of a sample of N = 306 participants doing a single negotiation working in pairs. They were assigned to 25 four-person groups and one three-person group. Given that SRM analyses require data from a minimum of four participants in each group, the group of three was discarded and a final N = 100 participants were included in the analyses below. From the unique 150 negotiations that resulted from these 100 individuals, data on outcomes (but not behavior) are excluded from eight of the negotiations—six due to impasses and two in which the counterparts reported discrepant outcomes. Although the participants were students, they had an average of 6 to 7 years of full-time work experience before returning to graduate school in management.

Round-Robin Interactions

In each group of four, members took turns engaging in one-on-one interactions with each other member of the group, using a round-robin format. As such, individuals are nested in varying dyads that are themselves fully nested within one group. For data analysis, we used the algorithms of the SRM, which are listed in Appendix B of Kenny (1994) and implemented in the R programming language. ² First, to test Hypothesis 1, the SRM allows us to measure actor effects, that is, how consistent individuals are in their signature of behaviors from one interaction to the next. This is similar to calculating an intraclass correlation to measure consistency. Statistically, the SRM calculates how much of a negotiator’s behavior in each separate interaction can be explained by their average behavior across all interactions. The SRM accounts for the interdependent nature of the data by subtracting person-level effects of partners to isolate the person-level effects attributable to each negotiator (i.e., the algorithms account for the missing data in that people do not negotiate with themselves). The actor effect can be interpreted akin to an R ² for individual-level effects. That is, if each negotiation encounter were in a single row, there would be 12 rows per round-robin of four people (six dyads and each dyad is represented twice, once for each partner), and the actor effects would be the random effect for each negotiator. Second, to test Hypothesis 2, the SRM allows us to measure partner effects, that is, how consistently an individual’s partners behave. That is, with the same format of data set in which each negotiation encounter was in a single row, the partner effects would be the random effect for the negotiation partner. Finally, to test Hypothesis 3, the SRM measures dyadic reciprocity between partners, that is, how much similarity there tends to be in behavior between the negotiator and counterpart. This is the intraclass correlation between the behavioral measures of each member of the dyad. Note that it is possible to apply the SRM to fully symmetric data such as rapport or synchrony, which do not differ across the two partners. That is, if A has high rapport with B, then B has equally high rapport with A. In such cases, actor and partner effects are still defined, but there is no output for dyadic reciprocity because it is a perfect value by design.

For supplemental analyses, we examine the associations between actor effects and negotiation performance. To do this, we outputted individual-level effects from the SRM. These are the equivalent of the average value of an individual’s behavior across their three separate encounters. As such, the values can be used in further analysis as a strictly individual-level variable.

Negotiation Simulations

As discussed above, personality theory emphasizes that the construct of personality involves consistency in an individual’s behavior when placed in the same situation. Along these lines, in attempting to reconcile decades of research showing relatively modest findings for behavioral consistency, Mischel and Shoda (1995) attributed this to unmeasured differences in the situations in which behavior has been assessed. Likewise, within negotiation in particular, Thompson (1990) argued that inconsistencies in past findings on personality can be attributed in part to methodological variations across studies. Different types of negotiations vary in their demands (Bazerman et al., 2000; DeRue et al., 2009; Walton & McKersie, 1965), and the extensive literature on Person × Situation interactions emphasizes the need to pay attention to situational consistency in attempting to draw out consistent individual-level effects. Accordingly, our participants engaged in a series of negotiation tasks that were as identical as possible in their objective structural characteristics. This feature of our design minimized the extent to which tasks differed in their situational constraints on behavior.

Accordingly, we created three negotiation scenarios that were all exactly parallel in the type of issues included and had the same underlying scoring system. They were adapted from the popular New Recruit exercise (Neale, 1997; Pinkley et al., 1994). Because Thompson (1990; Thompson et al., 2000) found that negotiators undergeneralize their learning from one negotiation setting to another, we kept the structure consistent while using three overtly different topics: the merger of two hotel chains, the purchase of a luxury car, and the rental of a home. Each case included five issues: two distributive issues (in which negotiators’ interests are fully opposed), one compatible issue (in which negotiators’ interest are fully aligned; Thompson & Hrebec, 1996), and two logrolling or trade-off issues (which vary in importance, so that there is opportunity to create value by trading off those of lower priority; Froman & Cohen, 1970; Pruitt, 1983). For each issue, the negotiators were given five options, a point value for each option, and a brief explanation to provide logic for the negotiator’s preference. A number of rules ensured that the three cases were identical: (a) all point values increased linearly, (b) every negotiation had the same maximum and minimum for distributive potential (4,700 and –200 points, respectively), (c) every negotiation had the same maximum and minimum for integrative potential (6,200 and 2,800 points, respectively), (d) the compatible issue was assigned smaller total values and values increased by smaller intervals, and (e) to provide as flexible as possible a benchmark for performance, an impasse was worth zero points. Note that all participants completed these three exercises in the same order, and any main effects for order were accounted for because the data were standardized within each role, as described below.

This task was assigned on the third meeting of the course and was the students’ first mixed-motive exercise of the semester. Each group took part in the study during a 90-min scheduled appointment, after all materials were available to prepare in advance. The 90-min session was split into three 30-min sessions. Within the four-person round-robin, pairs of partners interacted and then switched partners until, after the third session, each person had interacted with every other person. During the 30-min session, negotiators interacted for up to 15 min in a private room for each dyad, spent 10 min alone recording outcomes, and then had 5 min to take a break and switch rooms for the next interaction.

Outcome Measures

Participants reported their outcomes via survey immediately following each negotiation.

Coding Negotiation Frameworks

All negotiation exercises were video recorded, using a single camera fixed on the dyad. An external microphone increased the quality of the audio recordings. The recordings were transcribed into text and coded separately for each speaking utterance.

Negotiator Consistency (i.e., Actor Effects)

Hypothesis 1 examined the actor effects listed in Table 2. It is noteworthy that almost all behaviors showed significant or marginally significant consistency across the three interactions. As such, negotiators tended to act similarly in their micro-level behaviors from one encounter to the next. According to Kenny (1994), beyond statistical significance there is a benchmark of 5% to consider an effect meaningful, and by this benchmark many of the effects for behavioral consistency are substantial. For example, individual consistency from one encounter to the next explains 44% of the variance in asking questions, 35% of the variance in providing reasoning, and 26% of the variance in misleading.

Partner Consistency (i.e., Partner Effects)

In addition to negotiator behavioral consistency, Table 2 demonstrates evidence for partner behavioral consistency in the form of partner effects—albeit to a smaller degree. In support of Hypothesis 2, negotiators’ partners were consistent in terms of their development of rapport (34%) and their frequency of referring to alternatives (27%), making statements (23%), talkativeness (23%), making multiple-issue offers (22%), and making negations (21%).

Dyadic Reciprocity

Finally, we report the results for reciprocity. What the SRM model labels as dyadic reciprocity indicates the extent to which members of dyads uniquely reciprocate each other’s behavior. Consistent with Hypothesis 3, a number of behaviors were reciprocated. These include the use of nonfluencies (r = .81), providing reasoning (r = .60), talkativeness (r = .48), and making offers (multiple issue, r = .36 and single issue, r = .35), among others. Although this was not formally tested, by visual inspection the communication and procedural behaviors appeared to be reciprocated to a greater extent than did the competitive behaviors.

Associations With Negotiation Outcomes

As a supplementary analysis, we report the associations of the behavioral variables with negotiation outcomes. Table 3 presents correlations between each behavior and both objective value (total score, claiming value, and creating value) and subjective value (the negotiator’s own reports and partner reports). To conduct these analyses, we made use of individual-level variables that the SRM constructs. These variables reflect each participant’s personal average across their various partners, while accounting for interdependence of data points. A number of behaviors predicted objective scores, namely, misleading (r = .40), referring to fairness (r = .33), referring to alternatives (r = .28), using dictionary words (r = .28), using negations (r = .28), making multiple-issue offers (r = .25), and using the second-person pronoun you (r = .20). A number of behaviors also predicted subjective value, particularly for the subjective value of the partner. Most negotiation behaviors that correlated highly with claiming value also predicted lower subjective value for one’s partner.

OPEN IN VIEWER

Table 3.

Negotiation Process Measures and Outcomes (N = 150 dyads).

Criterion	Objective value			Subjective value
	Individual score	Claiming value	Creating value	Own	Partner
Negotiation behaviors
Communication behaviors
Asking questions	.14	.06	.18 †	−.21*	−.18 †
Making statements	.13	.10	.12	.01	−.10
Providing reactions	.16	.23*	−.11	−.18 †	−.48***
Providing reasoning	.11	.25*	−.26**	−.21*	−.50***
Procedural behaviors
Making single-issue offers	.13	.19 †	−.13	−.23*	−.42***
Making multiple-issue offers	.25*	.33***	−.12	−.12	−.37***
Procedural comments	.17 †	.12	.17 †	−.05	−.12
Referring to fairness	.33***	.29**	.18 †	−.1*	−.28**
Competitive behaviors
Referring to alternatives (BATNA)	.28**	.38***	−.15	−.15	−.54***
Referring to competitors	.12	.20*	−.17 †	.02	−.19 †
Referring to own status	.13	.24*	−.25*	−.23*	−.42***
Misleading	.40***	.48***	−.10	−.15	−.46***
Asking for sympathy	−.06	.07	−.28**	.06	−.29**
Relationship quality
Rapport	−.20 †	−.16	−.13	.15	.32**
Synchrony	−.13	−.12	−.03	.05	.12
Language use
Talkativeness (word count)	.15	.23*	−.13	−.21*	−.41***
Explicit confrontation (optimism)	−.08*	.03	−.23*	−.05	−.04
Dictionary words	.28**	.38***	−.15	−.29**	−.30**
Assents	−.14	−.15	.02	.15	.05
Negations	.28**	.29**	.05	−.15	−.37***
Positive emotion	−.07	−.12	.13	.22*	.21*
Negative emotion	−.01	.04	−.13	.07	−.06
Causation (cognitive mechanisms)	.13	.11	.09	−.15	−.22*
Insight	−.06	−.07	−.01	−.12	.20*
Tentative	−.17	−.11	−.14	−.06	.18 †
Possessive words (apostrophes)	.17 †	.18 †	−.02	−.26**	−.10
Discrepancies	.23*	.18 †	.18 †	.03	−.08
Certainty	−.01	.07	−.17 †	.04	.02
Nonfluencies	−.17 †	−.18 †	−.01	.17	.21*
Pronoun usage
I and self	.15	.10	.12	−.04	.04
We	−.04	−.08	.12	.15	.10
You	.20*	.19 †	.09	−.22*	−.24*
Verb tense
Past	.00	−.02	.04	−.02	.08
Present	.16	.17 †	.02	−.32**	−.09
Future	.05	.13	−.13	.00	−.24*

Note. 100 individuals. BATNA = Best Alternative to a Negotiated Agreement.

†

p < .10. *p < .05. **p < .01. ***p < .001.

Behaviors and Negotiation Performance

Analyses included the associations between all of the behavioral variables tested along with negotiation performance. Rather than make predictions for each of the 35 variables crossed with five factors of performance, we treated this as a supplemental analysis. Without formally testing hypotheses and controlling for the possibility of capitalizing on chance with this large number of coefficients, we note several broad trends post hoc based on visual inspection. In particular, there are a number of behaviors that predict higher scores—primarily through greater value claiming—and these behaviors appear to have a key attribute in common. Notably, they are all a matter of taking initiative and remaining actively engaged in the negotiation process. These include being more talkative, making more offers, referring to fairness, providing reasoning, and providing reactions. These also include competitive behaviors such as misleading and referring to one’s BATNA.

However, these highly activated behaviors tended in general to predict lower value creation. It is possible that the focus of these tactics on improving one’s own performance had the effect of crowding out room to engage in integrative discussion. We note that not all past findings regarding value creation are replicated in this study; notably, the negative effect of providing false or misleading information to the other party on value creation did not reach statistical significance. That said, as we discuss in the “Limitations” section, the distributive component of this exercise may have swamped the integrative component, in which case tests of coefficients for creating value may be underpowered.

Another trend is that many of the behaviors positively associated with value claiming—particularly the competitive behaviors—also are negatively associated with subjective value, especially the subjective value of the counterpart. This suggests caution for implementing “hardball” negotiation tactics. Such tactics may work in the short term, but backfire in the long term. We note that our research protocol does not test for the long-term quality of implementation or future interactions, both of which rely on goodwill to the extent that there is discretion involved, and so it may overestimate the potential value of highly competitive tactics.

Some smaller trends are also worth noting. For example, positivity (i.e., positive emotion) predicts greater subjective value, whereas negativity (i.e., negations) predicts lower subjective value. Talkativeness is evidently particularly disliked in the form of reducing counterparts’ subjective value considerably. This suggests the potential value for negotiators of being quiet enough to be an effective listener.

Individual Differences in Negotiation

These results suggest optimism for the study of individual differences in negotiation. The larger number of null and often inconsistent findings long associated with this question has led many theorists to speculate whether the pursuit itself should be abandoned (Terhune, 1970; Thompson, 1990)—in spite of strong intuitions by the general public that personality should matter. We drew upon theoretical traditions that consider individual differences to be consistent over time in an individual’s behaviors when placed in the same situation (e.g., Mischel & Shoda, 1995), and we took a different approach to this question than using personality questionnaires. In doing so, we find strong evidence that consistency over time exists, which is the very definition of personality. Such consistency suggests the potential for enduring dispositions to predict objective and subjective outcomes, ⁴ and so our study can be interpreted as contributing to the justification for continued interest in the topic. The results of this research suggest promising directions for future work.

New personality measures worth studying in the negotiations context could start with the list of behaviors that were highly consistent and that predicted differences in outcomes. This is a matter of coming “full circle” and letting the most relevant behaviors drive the choice of traits to feature in future work. On a post hoc basis, we note that there was an overarching theme among such behaviors of asserting oneself. As discussed in the section immediately above, the behaviors that appeared to improve objective performance were largely a matter of taking initiative and remaining actively engaged. A next frontier might be to map the behaviors from this study that proved to be behaviorally consistent onto granular-level personality traits that can be tested using questionnaires. This would mean moving beyond the higher level Big Five traits (Costa & McCrae, 1992), which have been extensively studied in the negotiations domain and which yield effect sizes that are close to zero for associations with performance variables (Sharma et al., 2013). Granular-level measures could be developed specifically for the negotiations domain (see, for example, negotiation self-efficacy; Sullivan et al., 2006), to supplement research that imports preexisting traits from general personality psychology. Given the theme of asserting oneself, some candidates of granular individual differences to feature may be along the lines of behavioral activation (Carver & White, 1994), assertiveness (Costa & McCrae, 1992), dominance (Wiggins, 1979), competing (Thomas & Kilmann, 1974), and unmitigated agency (Bakan, 1966; Helgeson & Fritz, 1999).

Limitations

This study has important limitations that need to be acknowledged. One limitation is the study’s laboratory nature using simulations, which is in keeping with the bulk of negotiations research. Scholars who study negotiation have tended to treat negotiations as decontextualized events, which is a choice reflected in the use of experimental designs (Barley, 1991). In addition, we attempted to be comprehensive in our selection of behaviors, but do not claim that this investigation exhausts all possible behaviors that are important in a negotiation framework. Another limitation is the sample size. Although the repeated nature of the round-robin design provided as much data as N = 300 participants in a standard design, and was more than sufficiently powered to test the study hypotheses using the SRM, for the supplemental analyses in which the individual was the unit-of-analysis, the sample was only N = 100 individuals. Another limitation is that the exercises had relatively low integrative potential for a mixed-motive scenario. Each contained one compatible issue and two issues that could be traded off against each other. The typical mixed-motive study uses eight issues (e.g., Neale, 1997), which offers more combinations that afford room for logrolling. Accordingly, the correlation between parties’ scores was r = –.70, which is largely distributive. As such, analyses of creating value could be underpowered.

Implications and Conclusion

We hope that the current study invigorates the further pursuit of personality effects in negotiation. In light of the meager findings for main effects in the body of past research, it was worthwhile to step back from this question. As much as our results suggest a strong role for individual differences in negotiation—given individual-level consistency in negotiators’ behavior—they also suggest the power of training. Systematic negotiation training has been demonstrated to be effective (Nadler et al., 2003), particularly using techniques of observational learning and analogical learning via comparative case studies (Loewenstein et al., 1999). We note that the behaviors that were the most predictive of negotiation effectiveness are within anyone’s grasp, such as making more multiple-issue offers, saying no more often, and referring to one’s alternatives or fairness principles when needed. Our focus on behavioral processes leaves open the door for negotiators to add effective tactics to their own repertoires without necessarily transforming the fundamentals of their personality.