## Introduction

People often view interpersonal, intergroup, and international relations as zero-sum, seeing one party’s gains as offset by other parties’ losses1,2,3,4,5. This belief can have broad implications, eroding interpersonal trust and increasing hostility, prejudice, misogyny, and anti-immigrant sentiment2,5,6,7,8,9,10. Even when people try to resolve their differences, expecting diametrically opposed outcomes can lead them to overlook mutually beneficial agreements, harm their relationships, and “leave money on the table”6,11,12. Regardless of whether a situation is truly zero-sum, viewing it as such clearly aggravates relationships between people, groups, and nations.

Research has assumed that people must engage in zero-sum interactions and has therefore focused on how they do so. Yet, many real-life situations allow people to choose whether they wish to interact with others, who they wish to interact with, and what level of interaction they seek. By focusing on how people behave in zero-sum situations, research has neglected the possibility that people may avoid such interactions altogether. Just as people sometimes avoid potentially beneficial economic games13,14,15, we suggest that they more generally avoid any situation that inversely links their and others’ outcomes. Thus, we raise a critical question: What happens when people can choose whether they wish to enter zero-sum situations?

We suggest that, when given the option to do so, people exhibit zero-sum aversionthe desire to avoid situations that are (or are believed to be) zero-sum. Importantly, we argue that this aversion is exhibited even when doing so is costly and both when situations are objectively zero-sum and when people subjectively see them as such. Rather than focus on people's reactions to existing outcomes of allocation decisions, we focus on their aversion to the process of such allocations. Independent of an allocation’s outcome (i.e., whether people receive more, less, or same as others), we suggest that people prefer to avoid processes that inversely links their and others’ outcomes.

Why do people exhibit zero-sum aversion? We suggest that people avoid zero-sum situations because they fear that these situations will elicit conflict and animosity. Zero-sum beliefs often make aggressiveness seem appropriate16,17,18 and people may therefore view zero-sum situations in general as prone to conflict. Such fear of conflict may arise in dyadic interpersonal and intergroup relations (e.g., expecting zero-sum resource allocation to elicit conflict between two individuals, groups, or nations) as well as in multi-party dynamics (e.g., expecting zero-sum organizational procedures to elicit conflict among many coworkers). Since people often withdraw from situations where they expect conflict15, and since such expectations often predict the actual conflicts people end-up experiencing19, they are likely to take their fear of conflict into consideration and avoid situations that are (or are believed to be) zero-sum.

Take, for example, the case of negotiations—a context with much potential for conflict. Decades of research examined how zero-sum beliefs affect trust, cooperativeness, and the “bottom line” in negotiations12. Yet, people often choose whether (and with whom) they wish to negotiate, leaving open the question of how seeing negotiations as zero-sum affects their willingness to do so. Although the propensity to initiate negotiations varies by gender, race, and other individual differences20,21,22, it is unclear if it is also affected by whether the negotiation itself is zero-sum. Thus, viewing negotiations as zero-sum may not only affect how people behave in them, but also whether they choose to negotiate in the first place.

This, of course, may be true of any situation people see as zero-sum. People may decline job opportunities where compensation is high but based on relative performance, entrepreneurs may avoid lucrative industries that require competing for market share, warring countries may avoid peace talks over seemingly zero-sum issues, and so forth. Fearing that zero-sum situations will foster conflict may lead people to avoid them, choosing less lucrative jobs, business endeavors, or peaceful courses of action.

We examine zero-sum aversion across thirteen main studies and eleven additional studies in the Supplementary Materials, documenting this phenomenon across various contexts that share an underlying psychological structure. First, we examine the effect of subjective zero-sum beliefs (Study 1A) and objective zero-sum situations (Study 1B), finding consistent evidence of zero-sum aversion in both real life and in hypothetical negotiations. Following, we examine the generalizability of zero-sum aversion in various real-life and hypothetical situations (Studies 2A-2C), which we then examine in a sequence of economic games involving real monetary outcomes (Studies 3A and 3B). Next, we test whether the fear that zero-sum situations will involve more conflict leads to zero-sum aversion (Studies 4A and 4B), whether the fear of conflict reduces people’s willingness to put other people in zero-sum situations (Study 4C), and whether manipulating this fear reduces zero-sum aversion (Studies 5A and 5B). Finally, we examine a real-world consequence of zero-sum aversion—increased wage requirements—among a sample of professionals (Study 6).

## Results

### Studies 1A and 1B

We begin by documenting zero-sum aversion in negotiations, a potentially costly, real-world context. Although negotiations are rarely zero-sum12, people often see them as such11,23,24. Yet, whereas past research focused on how zero-sum beliefs affect behaviors during negotiations, we examine whether such beliefs lead people to avoid negotiations altogether. Specifically, we examine how subjective beliefs about negotiations (whether negotiations are seen as zero-sum; Study 1A) and their objective structure (whether negotiations are indeed zero-sum; Study 1B) affect people’s propensity to engage in them and whether such zero-sum aversion is exhibited beyond any general individual preference for prosocial resource allocation.

#### Study 1A

We measured the prevalence of zero-sum beliefs and the propensity to initiate negotiations among a cohort of MBA students (N = 205). Compared to those with lower zero-sum beliefs, participants high in zero-sum beliefs exhibited substantial zero-sum aversion—believing that negotiations create harmful conflict (r(205) = 0.271, p < 0.001) and feeling apprehensive about initiating them (r(205) = 0.206, p = 0.003). In contrast, these participants were not more prone to recognize situations as negotiable (r(205) = − 0.023, p = 0.74) and actually felt more entitled to better outcomes (r(205) = 0.171, p = 0.014), suggesting that zero-sum beliefs were uniquely related to the propensity to initiate negotiations. Moreover, this relationship was exhibited even when controlling for the Big-Five personality traits (ß = 0.198, p = 0.005; Tables S1 and S2 in the SOM). Finally, we found that the belief that negotiations create harmful conflict mediated the effect of zero-sum beliefs on this avoidance (indirect effect: ß = − 0.280, SE = 0.069; 95% CI[− 0.413, − 0.146]; direct effect: ß = 0.008, SE = 0.069; 95% CI[− 0.127,0.144]). The more participants saw interpersonal relationships as zero-sum, the more they worried that negotiations create harmful conflict and, consequently, the more reluctant they felt about initiating them.

#### Study 1B

In Study 1B (N = 202), participants imagined working at an organization undergoing restructuring negotiations and were told that two issues were under consideration: a (zero-sum) resource allocation and a (non-zero-sum) workflow integration. They then indicated their willingness to represent their department in each negotiation.

We predicted that participants would avoid negotiations involving zero-sum issues. Indeed, participants exhibited zero-sum aversion, preferring the non-zero-sum integrative negotiation (M = 4.78, SD = 1.69) over the zero-sum distributive negotiation (M = 3.72, SD = 1.90), matched-pairs t(199) = 5.48, p < 0.0001, dz = 0.39.

Next, we examined whether zero-sum aversion was exhibited beyond individual differences in Social Value Orientation—a measure of social preferences in resource allocation25. Although social preferences moderated the aversion for inversely linking one’s own and others’ outcomes (F(2,195) = 3.31, p = 0.039), participants exhibited zero-sum aversion both when they were personally prone to favor prosocial allocations (Mdiff = 1.36, t(126) = 5.69, p < 0.0001, dz = 0.51) and when they were prone to favor individualistic allocations (Mdiff = 0.72, t(63) = 5.69, p = 0.047, dz = 0.25). While individual differences in social preferences matter (i.e., prosocial-leaning people exhibit somewhat stronger zero-sum aversion), the avoidance of zero-sum situations is exhibited beyond any such preferences.

### Studies 2A, 2B, and 2C

We next examine the generalizability of zero-sum aversion in controlled lab experiments involving market entry decisions (Study 2A), performance reviews (Study 2B), and real-effort tasks (Study 2C). In all studies, participants chose between a zero-sum option and a non-zero-sum option. We predicted that participants would exhibit zero-sum aversion, choosing to separate their and others’ outcomes over an option that inversely links them (i.e., zero-sum).

#### Study 2A

Study 2A (N = 100) examines zero-sum aversion in market entry decisions. We test whether people prefer to start businesses that expand the market (i.e., non-zero-sum) rather than capture existing market share (i.e., zero-sum) and whether this occurs even when success is more likely in the zero-sum option. Participants imagined choosing a location for a new business (a restaurant) between two options varying on several dimensions, including their potential to expand the market, with one location offering limited expansion and another location offering the potential for market expansion.

We predicted that participants would prefer to start a business that requires market expansion rather than a business that requires capturing existing market share, even when doing so can impede their success. Indeed, participants exhibited zero-sum aversion, choosing a location where customers would come from expanding the market (M = 5.18, SD = 1.53) over a location where they would have more customers but would need to attract them from other businesses (M = 4.00, SD = 1.54), matched-pairs t(99) = 4.33, p < 0.001, dz = 0.43. Thus, participants overwhelmingly opted for the riskier non-zero-sum option, choosing lower chances of success over a situation where others’ outcomes would come at their expense (and vice-versa).

#### Study 2B

Participants in Study 2B (N = 105) read about a company with two types of performance reviews, both of which offer equal chances of favorable evaluations: evaluating employees relative to their colleagues (i.e., zero-sum) or relative to absolute criteria (i.e., non-zero-sum). We predicted that participants would be averse to the zero-sum process, where one employee’s success comes at other employees’ expense. Indeed, participants significantly preferred to be evaluated by the non-zero-sum review (M = 5.68, SD = 1.63) over the zero-sum review (M = 2.95, SD = 1.87), matched-pairs t(104) = 9.22, p < 0.001, dz = 0.96. A similar pattern was exhibited in a forced-choice measure: 82% of participants (n = 86) preferred a non-zero-sum review, but only 18% (n = 19) preferred a zero-sum review, χ2(1,104) = 46.27, p < 0.001. Although both options offered equal chance of success, participants exhibited zero-sum aversion, choosing a (non-zero-sum) review that evaluates them relative to given performance criteria over a (zero-sum) review that evaluates them relative to their colleagues’ performance.

#### Study 2C

Study 2C (N = 199) examines zero-sum aversion in a real-effort task. As part of a workplace simulation, participants played the role of an intern in a magazine. In the Difficult Task condition, their job was to count the number of characters and spaces in several long and grammatically complex phrases. In the Simple Task condition, they simply counted the number of words in a few short phrases. Following, participants learned about a follow-up session involving similar tasks which may qualify them for a bonus. Participants then indicated their preference between attending a session in which the bonus allocation process would be zero-sum (i.e., based on relative performance) or non-zero-sum (i.e., based on absolute performance).

Participants disproportionately opted for the session where their bonuses would be independent from others’ bonuses: They significantly chose the non-zero-sum option (74.9%, n = 149) over the zero-sum option (25.1%, n = 50; χ2(1, 197) = 49.25, p < 0.001). This was true in both the Difficult Task condition (non-zero-sum: 75.8%, n = 72, zero-sum: 24.2%, n = 23, χ2(1, 93) = 24.04, p < 0.001) and the Simple Task condition (non-zero-sum: 74%, n = 77, zero-sum: 26%, n = 27, χ2(1, 102) = 25.27, p < 0.001), suggesting that the aversion to zero-sum reward allocations was unaffected by beliefs about the odds of outdoing others (χ2(1, 197) = 0.08, p = 0.776). Even when they thought they could complete the task faster and more accurately, participants chose to separate their and others’ outcomes.

### Studies 3A and 3B

Observing behaviors in controlled environments that model critical features of real-life situations is considered the “gold standard” for documenting robust psychological phenomena. Thus, Studies 3A and 3B examine zero-sum aversion in economic games involving real financial outcomes. Based on two pilot studies (Supplemental Studies S1 and S2 in the SOM), we gave participants a choice between two lotteries—one in which their payoffs would be inversely related to their counterpart’s payoffs (i.e., zero-sum) and one in which their payoffs would be independent from them (i.e., non-zero-sum). We predicted that participants would choose to bear real financial costs to avoid zero-sum situations.

#### Study 3A

In Study 3A (N = 207), participants were randomly assigned to play for either $1 or$5 and made a series of decisions between two lotteries with real monetary outcomes: a non-zero-sum option and a zero-sum option. In the non-zero-sum option, each player would enter a separate lottery, with 50% chance of gaining or losing $1 or$5 (depending on condition), guaranteeing that players’ payoffs are independent from each other. In the zero-sum option, both players would enter the same lottery, with 50% chance of gaining $1 or$5 (depending on condition) while their counterpart loses that same amount of money and 50% chance of losing $1 or$5 while their counterpart gains that money.

At this point, the experiment ended for participants who chose the zero-sum lottery. Participants who chose the non-zero-sum lottery decided between two additional options: the same non-zero-sum option from before (independently giving each player 50% chance of gaining or losing) and a new zero-sum option with better odds of winning (giving participants 60% chance of gaining while their counterpart loses and 40% chance of losing while their counterpart gains). Although this zero-sum option inversely links the players’ payoffs, it also offers participants better chances of winning than losing.

We repeated this procedure three times. Each time, participants who chose the safer zero-sum option ended the experiment. Those who chose the riskier non-zero-sum option saw two additional lotteries: the same non-zero-sum option (giving each player 50% chance of gaining or losing) and a new zero-sum option with even better odds of winning. This zero-sum option was, in sequence, a lottery with 70% chance of gaining (while their counterpart loses) and 30% chance of losing (while their counterpart gains), a lottery with 80% chance of gaining (while their counterpart loses) and 20% chance of losing (while their counterpart gains), and a lottery with 90% chance of gaining (while their counterpart loses) and 10% chance of losing (while their counterpart gains). The experiment ended after the 5th round.

Table 1 presents the results. First, we examined the percentage of participants who, in the first decision, chose the non-zero-sum option over the zero-sum option. Although both lotteries initially offered equal chance of winning, and despite playing for real money, participants were substantially more likely to choose the non-zero-sum lottery (78%, n = 161) over the zero-sum lottery (22%, n = 46), χ2(1,206) = 67.66, p < 0.001.

Of the 161 participants who chose the non-zero-sum lottery in the first choice, 74% (n = 118) still opted for it in the second choice rather than the zero-sum option with 10% better odds of winning, χ2(1,160) = 37.60, p < 0.001. Of these, 66% (n = 78) still chose the non-zero-sum lottery in the third choice rather than the zero-sum lottery that offered 20% better odds, χ2(1,117) = 12.46, p < 0.001. Of these 78 participants, 72% (n = 55) still chose the non-zero-sum lottery in the fourth choice rather than the zero-sum lottery that offered 30% better odds, χ2(1,75) = 15.76, p < 0.001. Finally, of the 55 participants who chose the non-zero-sum option in the previous four choices, 89% (n = 48) still preferred this lottery in the fifth (and final) choice over the zero-sum lottery that offered 40% better odds of winning, χ2(1,53) = 37.19, p < 0.001. Thus, participants exhibited zero-sum aversion even when winning the zero-sum lottery was 10%, 20%, 30% and even 40% more likely. On average, participants needed 19.9% better odds of winning to switch from the non-zero-sum lottery to the zero-sum lottery (a conservative estimate, treating participants who chose the non-zero-sum lottery in the final choice as likely to have switched to the zero-sum lottery if these chances rose by an additional 0.1%).

#### Study 3B

Participants in Study 3A made sequential choices, which may prompt strategic decision making such as attempting to be (or appear) consistent. To rule this out, we conducted a preregistered replication (N = 506) in which participants made a single choice between two lotteries: a non-zero-sum lottery and a zero-sum lottery with varying chances of winning.

### Study 4A

#### Participants

One hundred one U.S. residents were recruited from Amazon’s Mechanical Turk (Mage = 40.74; 45 females, 56 males; 70.3% White, 5.9% Black, 5.9% Hispanic, 13.9% Asian, 1% American Indian/Alaskan Native, 1% Native Hawaiian/Pacific Islander American, 2% Other), giving us 80% power to detect medium effects (d = 0.56) in an independent sample t-test.

#### Materials and procedure

Participants were randomly assigned to read about one of two performance reviews from Study 2B. In the zero-sum condition, they read about a procedure that reviews employees relative to each other, inversely linking their outcomes and rewarding only a set number of high-ranked employees. In the non-zero-sum condition, they read about a procedure that independently reviews employees and rewards those who outdo absolute criteria, but in which there is no guarantee that anyone would do so.

Participants indicated how much they believed the procedure would create conflict using five-items: “harmful conflict,” “friction,” and “animosity” between employees at the company, “pit employees” against each other, and put tension” on their relationships (1 = Not at all; 7 = Very much so; α = 0.98). Following, they reported how much they would want to be evaluated by the procedure (“To what extent would you like to be evaluated by the system that the company uses?” 1 = Not at all; 7 = Very much so).

### Study 4B

#### Participants

Two hundred fifty-one U.S. residents were recruited from Amazon’s Mechanical Turk (Mage = 40.15; 131 females, 118 males, 2 non-binary; 75.1% White, 6.1% Black, 5.8% Hispanic, 9.6% East Asian, 2.7% South Asian, < 1% Middle Eastern/Arabic, < 1% Other), giving us 80% power to detect medium effects (d = 0.36) in an independent sample t-test.

#### Materials and procedure

Participants were randomly assigned to read about one of two negotiations from Study 1B. In the zero-sum condition, they read that “the main issue to be discussed is resource allocation” and that representatives will need to “advocate for as many resources as possible” and “make sure that other departments don’t take resources away from [them].” In the non-zero-sum condition, participants read that “the main issue to be discussed is work integration” and that representatives will need to “advocate for integration that benefits everyone” and “make sure that [they] create as much value as possible.” Participants were warned in both conditions that the negotiations can become “heated” and that representing one’s division “is not going to be easy.” Following, participants indicated how much conflict the negotiations would create using the items from Study 4A (α = 0.94) and indicated their willingness to represent their department in them (“To what extent would you like to represent your department at these discussions?” 1 = Not at all; 7 = Very much so).

### Study 4C

#### Participants

Two hundred fifty-one U.S. residents were recruited from Amazon’s Mechanical Turk (Mage = 37.86; 105 females, 144 males, 2 prefer not to say; 72.5% White, 9.3% Black, 5.6% Hispanic, 7.1 East Asian, 3.0% South Asian, 1.1% American Indian/Alaskan Native, 1 < % Native Hawaiian/Pacific Islander American, < 1% Middle-Eastern/Arab, < 1% Other), giving us 80% power to detect medium effects (d = 0.36) in an independent sample t-test.

#### Materials and procedure

Participants were randomly assigned to read about one of two performance reviews from Study 2B. Unlike before, participants imagined they were managers at a company who were looking for new ways to evaluate their employees and were asked to consider the review process as a way of evaluating other people (i.e., their supervisees). In the zero-sum condition, participants read about a procedure that reviews employees relative to each other, inversely linking their outcomes and rewarding only a set number of high-ranked employees. In the non-zero-sum condition, participants read about a procedure that independently reviews employees and rewards those who outdo absolute criteria.

Participants indicated how much they believe the procedure would create conflict among their employees using five-items: “harmful conflict,” “friction,” and “animosity” between employees at the company, “pit employees” against each other, and put tension” on their relationships (1 = Not at all; 7 = Very much so; α = 0.98). Following, they reported whether they would choose to implement the review system for their employees (“How likely would you be to choose this system for your company?” 1 = Not at all likely; 7 = Very likely).

### Study 5A

#### Participants

Two hundred two U.S. residents were recruited from Amazon’s Mechanical Turk (Mage = 41.36; 123 females, 78 males; 80.4% White, 7.2% Black, 5.3% Hispanic, 3.8% East Asian, 1.4% South Asian, 1% Indigenous/Native American, < 1% Middle Eastern/Arabic, < 1% Other), giving us 80% power to detect small effects (dz = 0.20) in a matched-pairs design.

### Study 5B

#### Participants

Two hundred three U.S. residents were recruited from Amazon’s Mechanical Turk (Mage = 41.81; 107 females, 91 males, 4 Prefer not to say; 76.1% White, 6.6% Black, 7.5% Hispanic, 7.0% East Asian, < 1% South Asian, < 1% Indigenous/Native American, < 1% Middle Eastern/Arabic, < 1% Other), giving us 80% power to detect ratios as small as 1.32 in a chi-square design.

#### Materials and procedure

Following, participants chose between the two different negotiations in a trinary decision (“Which discussion would you choose to join?” The discussion about resource allocation, The discussion about workflow integration, or No Preference).

### Study 6

#### Participants

Three hundred five MBA students completed the study (Mage = 27.86; 117 females, 187 males, 1 prefer not to answer; 40% White/European American, 7% Black/African American, 11% Hispanic/Latin American, 33% Asian/Asian American/Pacific, 2% Middle Easter/Arab, 6% Other), giving us 80% power to detect small effects (f = 0.18) in a one-way omnibus ANOVA.

#### Materials and procedure

As part of a broader survey, participants read an excerpt from an ostensible job posting for Orion-Toren Consultants (OTC)—a “mid-sized, East coast-based boutique consulting company targeting recent MBA graduates”—and were randomly assigned to one of three conditions. In the zero-sum condition, participants read:

At Orion-Toren, we believe that only the best consultants can make it in the long run. As a consultant at OTC, you will undergo our proprietary professional-development program and will be evaluated based on how you perform relative to your incoming group of 8-10 other consultants. Consultants are ranked relative to each other, and receive constant guidance and evaluation, so that you will always know where you stand relative to your immediate peers. Needless to say, we expect all of our consultants to strive to beat their relative peer group. We take these performance evaluations seriously and use them to determine our consultants' compensation packages. Consultants who outdo their immediate peers are handsomely rewarded on a semi-annual basis.

In the non-zero-sum condition, participants read:

At Orion-Toren, we believe that only the best consultants can make it in the long run. As a consultant at OTC, you will undergo our proprietary professional-development program and will be evaluated based on how you perform relative to predetermined set benchmarks. Consultants are evaluated relative to these absolute benchmarks, and receive constant guidance and evaluation, so that you will always know where you stand relative to these organizational milestones. Needless to say, we expect all of our consultants to strive to beat their benchmarks. We take these performance evaluations seriously and use them to determine tour consultants' compensation packages. Consultants who outdo their absolute benchmarks are handsomely rewarded on a semi-annual basis.

Finally, in the control condition participants read:

At Orion-Toren, we believe that only the best consultants can make it in the long run. As a consultant at OTC, you will undergo our proprietary professional-development program and will be evaluated based on how you perform. Consultants receive constant guidance and evaluation, so that you will always know where you stand. We take these performance evaluations seriously and use them to determine tour consultants' compensation packages. Consultants who perform well are handsomely rewarded on a semi-annual basis.

After reading the job posting, participants were told that they learned through an acquaintance that the starting salary for consultant was, in the past, \$115,000 and indicated how much money they would ask for to work at this company (“What is the minimum salary you would be willing to accept to take this job?”). In addition, participants completed a single-item measures of self-esteem (“Please rate how much you agree with the following statement: I have high self esteem” 1 = strongly disagree, 7 = strongly agree) and an eight-item measure of their self-perceptions as better-than-average (“For the following questions, please rate yourself in comparison to the other students in this class: Decision-Making, Bargaining, Intelligence, Driving, Health, Charity, Physical Attractiveness, Good Friends”).