Zero-Sum Beliefs Across Age and Generations

Politics as well as economic policy often involves decisions about how resources, wealth, and opportunities are distributed across different sectors of society. Frequently, these issues are considered zero-sum, where one group's gain is seen as another group's loss. For example, elections are a zero-sum game. Only one person can become a president of a country; only one person can be elected as a senator for each position. Yet, our civic and political life is mostly comprised of situations that are not zero-sum. Economic systems are often dynamic, and the effects of policy decisions can create win-win situations where growth, innovation, and productivity benefit multiple groups simultaneously. While people often think about such issues as zero-sum, they rarely realize they are adopting such a zero-sum mindset (Andrews Fearon & Götz, 2024).

Three of the four experimental studies were preregistered, including their design, methodology, analysis, and exclusion criteria. The preregistrations can be accessed using the following links: Study 2 (https://aspredicted.org/Y9N_3D7), Study 3 (https://aspredicted.org/S6V_H5D), and Study 4 (https://aspredicted.org/JVL_NB8). De-identified data, analysis code, and research and additional online material are available on the Open Science Framework at https://osf.io/kv4pw/?view_only=e9f577673dab4402be5857870febb8fb.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this article.

Veronica Vazquez-Olivieri played a lead role in conceptualization, data curation, formal analysis, investigation, methodology, project administration, validation, visualization, writing-original draft, and writing-review and editing and a supporting role in funding acquisition. Tamar Kricheli-Katz played a supporting role in formal analysis, validation, and writing-review and editing. Boaz Keysar played a supporting role in conceptualization, funding acquisition, investigation, methodology, project administration, and writing-review and editing.

To understand whether a situation is zero-sum, one must consider the dynamics that give rise to it and its consequences. While people may view immigration as a zero-sum issue, others may see it as nonzero-sum by recognizing the mutual benefits immigration can offer. For example, many immigrants who enter the United States perform jobs that citizens do not want. Consequently, U.S. economists believe that an immigration surge typically allows businesses to meet the demand for their goods and services without competing for workers or raising wages (Alesina et al., 2023; Wile, 2024), resulting in a prospering economy with little to no increase in inflation rates. To view life as zero-sum is to believe that what you want and what others want are not compatible (Thompson & Hastie, 1990), that others gain at your expense (Roberts & Davidai, 2022), and that economic exchanges benefit sellers more than buyers (Johnson et al., 2022). This can have meaningful repercussions as zero-sum thinking is associated with lower life satisfaction, greater cynicism, and less trust for others and for societal institutions (Chernyak-Hai & Davidai, 2022; Davidai et al., 2022; Różycka-Tran et al., 2021).

Intuitively judging issues as zero-sum, even when there is potential for everyone to win, is a biased way of thinking that can lead to false judgments and poor decisions. For instance, people might see an issue as zero-sum depending on perceived challenge to the status quo (Davidai & Ongis, 2019). For example, when presented with a statement such as “The easier it is for black students to gain admission to college, the more it becomes difficult for white students to get admitted,” conservatives are more likely than liberals to see the issue as zero-sum. Yet liberals are more likely to see this issue as zero-sum if it is presented as “The easier it is for white students to gain admission to college, the more it becomes difficult for black students to get admitted” because it is maintaining the status quo. This is crucial for policymakers as regardless of the issue at hand, how they frame an issue can affect zero-sum thinking among constituents. Moreover, these beliefs affect behavior as those who believe success is zero-sum are less likely to help others succeed on their own (Chernyak-Hai & Davidai, 2022). Thus, believing others’ success comes at your expense has grave consequences for how people understand the world and behave in it.

Age and Zero-Sum Thinking

Here, we propose that zero-sum thinking may differ as one ages. Over time, individuals are able to acquire certain abilities based on experience. For example, older individuals outperform younger individuals in domain-specific tasks that benefit from experience such as chess (Roring & Charness, 2007) and crossword puzzles (Salthouse, 2004). Likewise, job performance improves with greater experience in high-complexity jobs (Sturman, 2003). Older adults are more holistic learners and better able to adaptively respond than younger adults in decision-making paradigms (Worthy et al., 2011), suggesting that older people may engage in more deliberation and wise reasoning. This has direct consequences for zero-sum thinking as considering the long-term effects of a situation and systematically processing its details can result in less zero-sum thinking (de Dreu et al., 2000). It is possible that through life experiences, older individuals learn to identify the win–win aspects of a situation and therefore endorse zero-sum beliefs less than younger individuals.

Potential Mechanisms

People’s motivations and how they evaluate situations can also change with age. Specifically, individuals in middle and late adulthood experience fewer negative emotions (Carstensen et al., 2011; Stone et al., 2010), and they also favor positive over negative information (Mather & Carstensen, 2005; Reed & Carstensen, 2012; Reed et al., 2014). This “positivity effect” leads older people to attend to and remember more positive information than negative information. It also relates to their pursuit of more emotionally meaningful goals that focus on savoring emotional experiences (Fung & Carstensen, 2003). Indeed, how much time one perceives to have left to live can shape their goals and preferences (Carstensen, 1992, 2006; Giasson et al., 2019). For example, older individuals who think they have 6 months left to live generate more emotionally meaningful bucket list goals than younger individuals under the same conditions (Chu et al., 2018). When having the choice between two advertisements, one related to exploration and the other to emotionally meaningful rewards, older individuals chose the latter and even remember the advertisement better compared to younger individuals (Fung & Carstensen, 2003). In this experimental paradigm, younger and older adults were presented with two advertisements that were identical except for the slogan (i.e., “Capture those special moments” or “Capture the unexplored world”) and then rated the advertisements on different dimensions followed by a recognition memory test. These types of paradigms allow researchers to test whether older individuals attend more to the positive and meaningful aspects of a situation. It is possible that this may result in seeing the win–win potential of situations. Hence, older individuals may be less zero-sum in their thinking than younger individuals. While zero-sum beliefs may not be inherently negative, we propose that, at times, these are seen as eliciting competition, threat, and cynicism and that having a more positive outlook might reduce zero-sum beliefs in certain contexts.

The degree to which one exhibits zero-sum thinking can also be the result of different intrapersonal and situational forces (Davidai & Tepper, 2023). When resources are scarce such as under unfavorable economic conditions or in times of financial vulnerability, zero-sum beliefs typically rise (Różycka-Tran et al., 2015). Older individuals compared to younger individuals typically have amassed wealth throughout their lifetime (Huggett, 1996) or created social connections that enhance their well-being (de Belvis et al., 2008; Steptoe & Fancourt, 2019). This can result in older people endorsing less zero-sum beliefs because they are less resource scarce.

Last, it is possible that an age-related reduction in zero-sum thinking is not the result of experience, environmental factors such as financial situation, or of how they attend to information. One of the most widely acknowledged psychological changes is that many, but not all, cognitive abilities decline with age. A global decline in cognitive functions is thought to be related to declines in working memory (Reuter-Lorenz et al., 2000) and specific aspects such as learning complex skills and sequences (Shea et al., 2006). Older individuals have more trouble recognizing deception and are less likely to detect lies compared to younger individuals (Ruffman et al., 2012). They are also generally more trusting than younger individuals and more likely to invest in others who have a negative reputation (Bailey et al., 2016). These types of errors might extend to zero-sum thinking, causing older individuals to endorse less zero-sum beliefs overall even when the situation is really zero-sum.

Research Overview

Zero-sum beliefs may manifest in two forms: (a) as a general mindset, resulting in a generalized view that one person’s gain comes at the other’s expense, and (b)as a situation-specific judgment, resulting in situation-dependent zero-sum beliefs (Davidai & Tepper, 2023). For example, one may conclude that a particular situation related to immigration is zero-sum, a specific judgment, but not endorse a general belief that others win at your expense across all circumstances and situations. These beliefs are not mutually exclusive, and our research systematically examines them to understand the relationships between them.

Using young and old participants, we consider differences in general beliefs, such as “If someone gets richer, it means that somebody else gets poorer,” and situation-specific judgments related to a workplace scenario. We determined our young and old age groups based on previous literature that identified young individuals as 17- to 35-year-olds (Kappes et al., 2020) or 18- to 35-/18- to 37-year-olds (Fung & Carstensen, 2003) and old individuals as ages 65–85 (Kappes et al., 2020) or 65–85/55–83 (Fung & Carstensen, 2003). We focus on a particular workplace situation that is inherently nonzero-sum, where employees are evaluated compared to expectations and not to each other. We examine if general zero-sum thinking leads people to perceive this situation in a zero-sum way. In other words, to perceive that an employee can only succeed at the expense of others.

In examining both types of zero-sum beliefs, we also consider the mechanisms that may explain differences in zero-sum thinking with age. For instance, given that individuals become more positive with age (Mather & Carstensen, 2005), older individuals may be less zero-sum because they are more positive. Given that zero-sum beliefs rise when resources are perceived to be scarce (Davidai & Tepper, 2023; Meegan, 2010), older individuals may also be less zero-sum because they perceive less resource scarcity.

When comparing young and old people synchronously, differences could reflect the impact of aging, but they could also reflect generational differences, or both. In order to separate these effects, we then used the World Values Survey (WVS), which captures “social, political, economic, religious, and cultural values of people in the world” (World Values Survey, n.d.). The survey included a question that could reflect zero-sum belief, so we capitalized on it to compare such thinking at two points in time, separated by 2 decades. This allowed us to investigate the contribution of age separately from the contribution of belonging to a particular generation. In addition, given that the surveys were conducted around the world with more than 200 thousand people, they allowed us to test the generalizability of the results beyond a Western, educated, industrialized, rich, and democratic population (Henrich et al., 2010).

Transparency and Openness

Three of the four experimental studies were preregistered, including their design, methodology, analysis, and exclusion criteria. The preregistrations can be accessed using the following links: Study 2 (https://aspredicted.org/Y9N_3D7), Study 3 (https://aspredicted.org/S6V_H5D), and Study 4 (https://aspredicted.org/JVL_NB8). De-identified data, analysis code, and research and additional online material are available at https://osf.io/kv4pw/?view_only=e9f577673dab4402be5857870febb8fb.

Study 1: Documenting the Age Effect

Study 1 examines how zero-sum beliefs differ as a function of age. Specifically, we examine general zero-sum beliefs that impact how people view the world, as in “Successes of some people are usually failures of others.” We also consider situation-specific zero-sum judgments by presenting a workplace situation where employees in a division are evaluated based on their productivity compared to expectations. Individuals may erroneously perceive that for one employee to succeed, another must fail, even though employees’ compensation is not determined by comparison to others’ performance.

Method

Participants

An a priori power analysis, conducted using G*Power3.1 (Faul et al., 2009), for a two-tailed t test assuming a small effect size (d = .20) and a minimum power of .8 indicated a minimum sample of 788 participants. Therefore, we recruited 788 individuals on Prolific. To be included, participants had to successfully complete an attention check and be native English speakers. There was a 14% rate of attention check failure, resulting in a final sample of 668 participants ($N_{\text{Young}} = 337$, $N_{\text{Old}} = 331$). The young group was between the ages of 18 and 30, the mean age was 24.9 years, 164 (48.7%) identified as females, most identified as White (216; 64.1%), and the majority had at least a trade school certification (210; 64.1%). The old group was between the ages of 65 and 80, the mean age was 69.3 years, 159 (48%) identified as females, most identified as White (321; 97%), and the majority had at least a trade school certification (205; 61.9%). Participants were asked about their employment status, political beliefs, if they were retired, if they had children, their net worth, and their income for the past 12 months (see the Supplemental Materials).

Materials and Procedure

Participants were told they would be completing a series of three tasks where they would be asked to make decisions. In the first task, participants were presented with eight randomized statements from the Belief in a Zero-Sum Game scale (Różycka-Tran et al., 2015). For example, they rated arguments such as this on a 7-point scale: “If someone gets richer, it means that somebody else gets poorer,” with 1 as completely disagree and 7 as completely agree. We averaged the items to create a measure of general zero-sum belief.

Second, we considered the possibility that with age, people might encounter more situations that are not zero-sum, leading them to endorse less zero-sum beliefs. To evaluate this, we asked them to generate counterexamples to the following statement: “In order for someone to get much richer, other people must become a bit poorer.” Participants were given examples of what would qualify as a good counterexample and were incentivized with $0.25 bonus per good example. They were also warned against the use of artificial intelligence (AI) by being informed their responses would be submitted to an AI detector and they would receive no bonus if any response was determined to be generated by AI. Participants were given 5 min to complete the task and were allowed to proceed with the experiment after 2 min had elapsed.

Two research assistants who were blind to the age of the participant coded the examples. For each response generated by the participant, the research assistants evaluated if it was a good counterexample to the statement (0 = bad example, 1 = good example). To help the raters understand the scope of each category, we provided examples of coded responses and emphasized that the list was not exhaustive. For instance, lotteries or inheritance (e.g., “Lottery winners do not take from others”) or consequences from wealth (e.g., “If someone gets richer, many people around benefit from it, his loved ones for example”) were considered bad responses as these did not explain how one’s gain did not come at another’s expense. Conversely, good examples could stem from innovations (e.g., “When something is created for the first time and fills a niche, nobody losses out and everyone gains from it”) or financial opportunities (e.g., “When people invest in the stock market, that isn’t a zero-sum game. With the steady increase on average of the market over time, in the long run most all investors can benefit.”). To ensure the raters understood the instructions, they reviewed the participants’ stimuli and coded 20 practice responses that were not collected in the experiment. After the initial round of coding was done independently, the raters convened to resolve any discrepancies in their reports.

Participants then read about the following workplace situation adapted from Davidai et al. (2022) and were asked to evaluate the extent to which it was zero-sum.

Imagine that you work for a company that is well known but does not have a good work–life balance. Employees in the company typically feel that they’re not adequately compensated, that they lack opportunities to learn new skills, and that the work is rarely challenging and meaningful. This company uses the “Rated Success System” for compensation. This is how the system works: Every year, all employees in a division are evaluated based on their productivity as compared to expectations. Those who receive very high ratings get bonuses, extra vacation days, and even promotions. Those who receive very low ratings do not receive a bonus and their contracts are sometimes “terminated.” Of course, the higher evaluation people receive, the more they are considered successful in the organization, and no one wants to receive a low evaluation.

This situation is inherently nonzero-sum because employees are compared to expectations and not to each other. Participants were then asked, “Zero sum is a situation where one person’s gain is equivalent to another person’s loss. To what extent do you perceive this evaluation system as zero-sum?” (1 = not at all, 7 = very much so). Participants were also asked to explain their selection on the scale, “Please justify your selection on the scale above by explaining why.” Two research assistants blind to the age of the participant and their rating coded the explanations provided (0 = nonzero-sum, 1 = zero-sum, 999 = nondiagnostic). The coding evaluated whether the participant’s justification captured zero-sum thinking (e.g., “To get a high ranking, someone else must get a lower ranking, therefore getting less bonuses, vacations, etc.”), nonzero-sum thinking (e.g., “If all employees scored highly, they would all get bonuses”), or nondiagnostic (e.g., “The person with a high gain is far better off than the other person who gains nothing”). Last, participants completed an attention check and a battery of demographic measures.

Results

Younger participants exhibited more zero-sum beliefs than older participants both when evaluating general statements and a specific workplace situation. For the general beliefs, younger participants ($M = 4.39$, $SD = .95$) endorsed significantly more general zero-sum beliefs than older participants ($M = 3.52$, $SD = 1.05$), $t(656.8) = -11.37$, 95% CI $[-1.03, -.73]$, $p < .001$, Cohen’s $d = -.88$, 95% CI $[-1.04, -.72]$ (Figure 1). An item analysis demonstrated that younger participants endorsed zero-sum beliefs more than older participants for all items, with differences ranging from .36 to 1.18. This suggests that older individuals hold significantly less general zero-sum beliefs than younger individuals.

While the scenario presented a situation that was inherently nonzero-sum, younger participants ($M = 3.85$, $SD = 1.98$) judged it as significantly more zero-sum than older participants ($M = 3.53$, $SD = 2.05$), $t(664.18) = -2.03$, 95% CI $[-.62, -.01]$, $p < .05$, Cohen’s $d = -.16$, 95% CI $[-.31, -.01]$. Next, we considered the explanations participants provided for their ratings. To assess the agreement between the three raters on their independent coding, we computed Cohen’s kappa unweighted and found substantial agreement (Cohen’s $\kappa = .72$, $z = 26.1$, $p < .001$). Excluding responses that were nondiagnostic, we performed a chi-square test to determine if the distribution of explanations that were coded as zero-sum and as nonzero-sum differed by age. Indeed, the proportion of participants who provided reasonings containing zero-sum beliefs was significantly different by age, $\chi^2(1, N = 533) = 9.65$, $p = .002$, 95% CI $[-.22, -.05]$. Of the young participants’ explanations, 45% exhibited zero-sum thinking, whereas only 32% of old participants’ explanations did. Taken together, this suggests that younger individuals’ belief in zero-sum biases them to judge nonzero-sum situations as significantly more zero-sum than older individuals do. This is evident from both their ratings and the justifications they provided.

The qualitative analysis for the counter-example generation task showed no difference by age. While there was substantial agreement between raters both when evaluating each counterexample generated by a participant (Cohen’s $\kappa = .84$, $z = 34$, $p < .001$) and when considering the total count of good examples per participant (Cohen’s $\kappa = .77$, $z = 32.3$, $p < .001$), there was no difference in the number of examples generated as a function of age. On average, old participants generated 0.94 good responses and young participants 0.98 responses, $t(659.78) = -.50$, 95% CI $[-.21, .12]$, $p = .62$. There was an equal percentage (74%) of old and young participants who scored between zero and one good response. Thus, the majority of participants across age groups generated up to one good response. To evaluate if this was the result of the coding scheme being too strict, we also evaluated if the number of examples generated differed by age, regardless of how good the examples were. On average, old and young participants generated the same amount, about 2.5 examples.

Next, we conducted simple linear regressions to evaluate the extent to which situation-specific judgments and age (young or old) could predict general zero-sum belief scores. Indeed, we find a significant model, $F(2, 665) = 77.23, p < .001, R^2 = 0.19$. Both situation-specific judgments ($\beta = 0.09, p < .001$) and age ($\beta = 0.85, p < .001$) were significant predictors of general zero-sum beliefs. Taken together, this suggests that an individual's zero-sum judgment of a specific scenario and their age play an important role in shaping their general zero-sum thinking.

It is possible that environmental factors, such as a person's financial situation, influence the degree of their zero-sum thinking, so we examined the relationship between demographic measures and general zero-sum beliefs. We conducted exploratory moderation analysis to test whether income, education, gender, political beliefs, employment, net worth, retirement, or having children impacted the effect of age on general zero-sum beliefs. Net worth was the only variable that significantly interacted with age. There was a significant main effect of age, $F(13, 582) = 11.37, p < .001, \eta_p^2 = .18$, and a significant interaction between age and net worth at the level of $100,000–$149,000 ($b = .02$), $t(582) = 2.08, p = .04, \eta_p^2 = .01$, and of $150,000+ ($b = .02$), $t(582) = 2.22, p = .03, \eta_p^2 = .01$. To better examine the relationship between age (young and old), net worth, and general zero-sum beliefs, we conducted an analysis of variance (ANOVA) by collapsing the seven categories of net worth into three levels: low ($0–24,999), medium ($25,000–99,999), and high ($100,000+). We opted to proceed with an ANOVA given that our independent variable and moderator were both categorical; hence, we wanted to test their interaction. We found a significant main effect of age, $F(1, 590) = 126.98, p < .001, \eta^2 = .17, 95\% \text{ CI } [.13, 1]$; a main effect of net worth, $F(2, 590) = 3.80, p = .02, \eta^2 = .01, 95\% \text{ CI } [.0, 1]$; and a significant interaction between the two, $F(2, 590) = 3.19, p = .04, \eta^2 = .01, 95\% \text{ CI } [.0, 1]$. Bonferroni post hoc tests revealed that young individuals high in net worth ($M = 3.57, SD = 1.06$) were significantly less zero-sum than their counterparts at low ($M = 4.45, SD = .93$) and medium net worth ($M = 4.49, SD = .86$). Low and medium net worth young individuals are significantly more zero-sum than old individuals from all net worth levels (old low: $M = 3.47, SD = 1.06$; old medium: $M = 3.64, SD = 1.03$; old high: $M = 3.40, SD = 1.10$). This suggests that wealthy young and older people hold similar (and low) zero-sum beliefs. The interesting finding is that we still see a significant reduction in zero-sum beliefs between ages for low and medium net worth individuals (Figure 2). A reduction in zero-sum beliefs with age as a result of resource scarcity (net worth) is partly supported by the data because while resource scarcity may be driving zero-sum beliefs among young individuals, scarcity does not impact older individuals to the same extent. It is possible more than one mechanism is driving differences in zero-sum beliefs with age, and we do not consider these two mechanisms to be mutually exclusive.

Discussion

We find that older individuals hold significantly less general and situation-specific zero-sum beliefs compared to younger individuals. We also find some support for the idea that scarcity plays a role in the age effect, as wealth interacts with it: People with low and medium wealth show less zero-sum thinking with age. Yet people who do not experience scarcity, the high net worth individuals, show no age effect. This is consistent with a resource scarcity theory because it shows that scarcity is necessary for the age effect (Różycka-Tran et al., 2015). Last, we find evidence that situation-specific judgments and age play an important role in shaping people's general zero-sum thinking.

Study 2: A Replication Controlling for Executive Function

Study 2 has three goals. First, it provides a preregistered replication of the findings in Study 1. Second, given that executive functions decline with age (Fjell et al., 2017; Zelazo et al., 2004), we aimed to evaluate the role that executive functions might play in zero-sum thinking. Finally, in addition to the workplace situation that is inherently nonzero-sum, we included a zero-sum version to evaluate if age differences extend to such situations as well.

Method

Participants

An a priori power analysis, conducted using G*Power3.1 (Faul et al., 2009), for a two-way ANOVA for four groups (old zero-sum, young zero-sum, old nonzero-sum, young nonzero-sum) assuming a medium effect size $d = .20$, power of .80, yielded 416 participants. The final sample consisted of 416 participants, 208 young (18–35) and 208 old (65–80) that met the same inclusion criteria as Study 1. Different from Study 1, to be included, participants had to be born in the United States and currently reside in the United States. Participants who failed the attention check were replaced. In the young group ($N = 208$), the mean age was 25.82 years, 111 (53.3%) identified as males, 116 (55.8%) identified as White, and 120 (57.7%) had at least a trade school certification. In the old group ($N = 208$), the mean age was 69.18 years, there was an equal split across genders (105, 50.5%, identified as females; 103, 49.5%, identified as males), 193 (92.8%) identified as White, and 158 (75.9%) had at least a trade school certification. The same set of demographics as Study 1 was also examined (see the Supplemental Materials).

To assess executive functioning, participants completed the Trail-Making Test (TMT; Reitan & Wolfson, 1995), and to assess memory and recognition, participants completed a Word List Paradigm. Both tests were administered online through Cognition Lab. TMT consists of two parts: TMT-A is a measure of processing speed, in which participants are asked to connect 25 encircled numbers sequentially, and in TMT-B, participants must alternate between numbers and letters (i.e., 1, A, 2, B, 3, C). The Word List task presents 15 words and 30 word recall tests where participants must identify each word as new or old. We included a measure of executive functioning to evaluate whether it played a role in the observed age effect. Specifically, we opted for the TMT and Word List Paradigm, as these serve to measure attention, working memory, and memory.

In terms of executive functioning, we compared time on task on both TMT parts across ages and found that indeed, young participants outperform older participants—TMT-A: $M_{\text{young}} = 20.92$ s, $SD_{\text{young}} = 4.37$ s, versus $M_{\text{old}} = 30.61$ s, $SD_{\text{old}} = 7.55$ s, $t(331.62) = 16.02$, 95% CI [8.50, 10.88], $p < .001$, Cohen's $d = 1.57$, 95% CI [1.35, 1.79]; TMT-B: $M_{\text{young}} = 43.14$ s, $SD_{\text{young}} = 15.66$ s, versus $M_{\text{old}} = 64.06$ s, $SD_{\text{old}} = 31.65$ s, $t(302.66) = 8.54$, 95% CI [16.10, 25.74], $p < .001$, Cohen's $d = 0.84$, 95% CI [0.64, 1.04]. For word recall, the two age groups did not differ on the number of words recalled: $M_{\text{young}} = 12.26$, $SD_{\text{young}} = 2.23$, versus $M_{\text{old}} = 12.35$, $SD_{\text{old}} = 2.14$, $t(413.33) = 0.45$, 95% CI [-0.33, 0.52], $p = .65$, Cohen's $d = .04$, 95% CI [-.15, .24]). These scores are consistent with TMT normative data for these ages, suggesting our sample is representative (Tombaugh, 2004). We utilized the cognitive measures as a basis to exclude and replace participants who scored worse than 3 standard deviations away from the mean.

Materials and Procedure

We used the situation-specific judgment and general belief items from Study 1. For the workplace situation, we created a zero-sum scenario version based on the nonzero-sum scenario. Whereas, in the nonzero-sum scenario, the company used a “Rated Success System” where employees were rated and compared to expectations, in the zero-sum scenario, the company used a “Ranked Success System” where employees were ranked and compared to each other. Additionally, because all items of the Zero-Sum Game scale (Różycka-Tran et al., 2015) showed an age difference in Study 1, we simplified the task and presented only three items to measure general zero-sum beliefs: “If someone gets richer, it means that somebody else gets poorer,” “Life is so devised that when somebody gains, others have to lose,” and “When some people are getting poorer, it means that other people are getting richer.”

Within age groups, participants were randomly assigned to either the nonzero-sum or the zero-sum scenario. In both scenarios, participants read the situation, completed the main dependent variable (i.e., “Zero sum is a situation where for a person to gain, another needs to lose. To what extent do you perceive this evaluation system as zero-sum?”; 1 = not at all to 7 = very much so), and provided a justification for their ratings. Participants then completed an attention check, three general zero-sum belief items, demographics, and, last, cognitive measures (TMT and Word Lists).

Results

Replicating Study 1, participants in the young group ($M = 4.50$, $SD = 2.06$) judged the nonzero-sum scenario as significantly more zero-sum than older participants ($M = 3.71$, $SD = 2.06$), $t(206.98) = -2.76$, 95% CI [-1.35, -.22], $p = .006$, Cohen's $d = -.38$, 95% CI [-.65, -.11]. Interestingly, participants from both age groups judged the zero-sum scenario comparably ($M_{\text{Young}} = 5.89$, $SD_{\text{Young}} = 1.56$; $M_{\text{Old}} = 5.68$, $SD_{\text{Old}} = 1.72$). There was a significant main effect of scenario type, $F(1, 412) = 84.57$, $p < .001$, $\eta^2 = .17$, 95% CI [.12, 1], and of age, $F(1, 412) = 7.52$, $p = .006$, $\eta^2 = .01$, 95% CI [0, 1], but no significant interaction between the two ($p = .12$; Figure 3).

As in Study 1, the justifications participants provided for their ratings were coded (0 = nonzero-sum, 1 = zero-sum, 999 = non-diagnostic). Three research assistants blinded to the age of the participant, scenario type, and their rating coded the explanations. To help raters understand the categories and ensure they understood the instructions, they coded 20 practice items that were not collected from the experiment and received feedback. After the initial round of coding that was done independently, the raters convened to resolve any discrepancies in their reports. To assess the agreement between the three raters on their independent coding, we computed Light's kappa and found good and substantial agreement (Light's $\kappa = .72$, $z = 12.1$, $p < .001$). Excluding responses that were coded nondiagnostic, we performed a chi-square test to determine if the amount of zero-sum and nonzero-sum responses differed by age and scenario type. For the nonzero-sum scenario, 60% of young participants provided zero-sum explanations, whereas only 49% of older participants did ($p = .21$). When evaluating the zero-sum scenario, 93% of younger participants and 87% of older participants provided zero-sum justifications ($p = .29$).

By collapsing the three zero-sum scale items into one average score, we again found that younger participants ($M = 4.60$, $SD = 1.41$) held significantly higher general zero-sum beliefs than older participants ($M = 3.50$, $SD = 1.61$), $t(406.85) = -7.39$, 95% CI [-1.39, -0.80], $p < .001$, Cohen's $d = -.72$, 95% CI [-.92, -.53]. Moreover, there was a significant age difference for each item in the scale ($ps < .001$), suggesting there is no one item driving this effect.

Next, as in Study 1, we conducted a linear regression to evaluate the extent to which situation-specific judgments (nonzero-sum scenario) and age (young or old) could predict general zero-sum belief scores, while accounting for executive functioning (the relative performance on both TMT tasks) as an additional predictor. We again find a significant model, $F(3, 205) = 35.11, p < .001, R^2 = 0.34$. Both situation-specific judgments ($\beta = 0.33, p < .001$) and age ($\beta = 1.01, p < .001$) were significant predictors of general zero-sum beliefs, while executive functioning was not ($\beta = 0.06, p = .63$). Consistent with our theory, we validate that executive functioning is not a significant predictor of zero-sum beliefs, suggesting a change in executive functioning cannot explain differences in zero-sum beliefs among young and older participants.

Like in Study 1, we examined the relationship between demographic scores and general zero-sum beliefs by conducting moderation analyses. The only variable that significantly interacted with zero-sum beliefs was employment. Specifically, we centered the outcome variable, zero-sum scale belief scores, and tested the interaction between employment and age, which was statistically significant for part-time employment ($B = -1.33, 95\% \text{ CI } [0.01, 1.20], p < .01$). We proceeded by dummy-coding part-time employment and found the interaction effect between working part-time and age was no longer significant, suggesting that employment is not impacting zero-sum scores.

Next, we explored the impact of executive functioning on zero-sum beliefs. On the one hand, the executive functioning decline that comes with age may contribute to a reduction in zero-sum beliefs through reduced systematic processing of the information. On the other hand, reductions in executive functioning may not adversely impact zero-sum beliefs, as lack of deliberation contributes to greater, not lower, zero-sum beliefs (Davidai & Tepper, 2023). Therefore, we considered the relationship between age, executive functioning, and zero-sum beliefs. To do so, we conducted correlation analysis by age on general zero-sum beliefs and executive functioning and found no significant relationships (Table 1).

We also examined situation-specific zero-sum judgments by performing a median split on the cognitive measures by age. Interestingly, we found that young participants who took longer on both TMT-A ($M_{\text{TMT-A-LOW}} = 5.07$) and TMT-B ($M_{\text{TMT-B-LOW}} = 5.02$; low performers) judged the nonzero-sum scenario as significantly more zero-sum than high performers on this task ($M_{\text{TMT-A-HR(H)}} = 3.86, M_{\text{TMT-B-HR(H)}} = 3.96$): TMT-A, $t(97.87) = -3.11, 95\% \text{ CI } [-1.99, -.44], p = .002$, Cohen's $d = -.61, 95\% \text{ CI } [-1.01, -.22]$; TMT-B, $t(101.42) = -2.69, 95\% \text{ CI } [-1.83, -.27], p < .01$, Cohen's $d = -.53, 95\% \text{ CI } [-.92, -.14]$ (Figure 4). In other words, high performers on these two cognitive measures rated the nonzero-sum scenario as significantly less zero-sum, lending support to the idea that increased processing results in reduced zero-sum thinking. Last, we examined if cognitive performance predicted our observed effect and found no significant evidence of this.

Age	Cognitive measure	t	df	p	r	95% CI
Young	TMT-A	.39	206	.70	.03	[-.11, .16]
	TMT-B	1.37	206	.17	.10	[-.04, .23]
	Word List recall	.69	206	.49	.05	[-.09, .18]
Old	TMT-A	.24	206	.81	.02	[-.12, .15]
	TMT-B	-.54	206	.60	-.04	[-.17, .10]
	Word List recall	-.86	206	.39	-.06	[-.19, .08]

Discussion

Study 2 replicated the pattern from Study 1, showing that old individuals judge nonzero-sum situations as less zero-sum than young people. They also endorsed general zero-sum beliefs less than young people did. Replicating Study 1, we find that both situation-specific judgments and age predict general zero-sum beliefs, whereas executive functioning does not. Given that the age effect persisted after controlling for executive functioning, cognitive decline cannot explain the results.

Study 3: A Powered Replication and Control

We had two goals in Study 3. First, Study 3 attempted a powered replication of results from Studies 1 and 2. The second goal was to provide control for the general beliefs. Older adults were indeed less likely to endorse the general belief that reflects that one person's gain must mean another's loss, but in principle, this could reflect a tendency to provide lower ratings independent of the content of the belief. To control for that, we included items that are about beliefs that are irrelevant to zero-sum: for example, the idea that 'Immigration has advantages and disadvantages.' If the age difference is specific to zero-sum thinking, then older adults will endorse the zero-sum items less than young people, but this will not happen with the control items.

Method

Participants

We consulted the same power analysis as Study 1 and replaced participants who did not meet the inclusion criteria. The final sample consisted of 788 participants, 394 young (20–35) and 394 old (65–80) that met the same inclusion criteria as Study 2. In the young group, the mean age was 28.83 years, there was an equal split across genders (192, 48.7%, identified as females; 188, 47.7%, identified as males), 254 (64.5%) identified as White, and 286 (72.6%) had at least a trade school certification. In the old group, the mean age was 68.54 years, 235 (58.6%) identified as females, and a majority identified as White (347; 88.07%) and had at least a trade school certification (245; 62.2%). The same set of demographics as Study 1 was also examined (see the Supplemental Materials).

We presented the same executive functioning measures from Study 2. We compared time on task on both TMT parts across ages and found that young participants outperform older participants—TMT-A: $M_{\text{young}} = 20.92$ s, $SD_{\text{young}} = 8.50$ s, versus $M_{\text{old}} = 30.93$ s, $SD_{\text{old}} = 7.81$ s, $t(780.51) = 17.19$, 95% CI [8.86, 11.14], $p < .001$, Cohen's $d = 1.22$, 95% CI [1.07, 1.38]; TMT-B: $M_{\text{young}} = 41.36$ s, $SD_{\text{young}} = 14.71$ s, versus $M_{\text{old}} = 62.25$ s, $SD_{\text{old}} = 26.1$, $t(619.91) = 13.84$, 95% CI [17.93, 23.86], $p < .001$, Cohen's $d = .99$, 95% CI [.84, 1.13]. For word recall, both age groups correctly recalled a similar amount of words ($M_{\text{young}} = 11.96$, $SD_{\text{young}} = 2.17$, vs. $M_{\text{old}} = 11.99$, $SD_{\text{old}} = 2.37$), $t(780.38) = .19$, 95% CI [-.29, .35], $p = .85$, Cohen's $d = .01$, 95% CI [-.13, .15]. While we utilized the cognitive measures as a basis to exclude and replace participants who scored worse than 3 standard deviations away from the mean, our scores are also consistent with TMT normative data for these ages, suggesting our sample is representative (Tombaugh, 2004).

Materials and Procedure

Study 3 used the same materials as Study 2 with the following distinctions. First, we only presented the nonzero-sum scenario. Second, we added two control items when measuring general zero-sum beliefs (“Immigration has advantages and disadvantages,” 1 = strongly disagree to 7 = strongly agree; “Government policies often have both supporters and critics with differing perspectives,” 1 = strongly disagree to 7 = strongly agree). The three experimental items and the two controls were presented in a random order.

Additionally, to identify potential mechanisms for the reduction in zero-sum beliefs, we measured positive thinking and perceived scarcity. Positive thinking, defined as the tendency to view life and experiences with a positive outlook, was measured using the P-Scale (Caprara et al., 2012; e.g., “I look forward to the future with hope and enthusiasm”). Given that individuals become more positive with age (positivity effect; Mather & Carstensen, 2005), older individuals may be less zero-sum because they are more positive. As a result, older individuals may be quicker to recognize that the workplace situation is not zero-sum and to generally endorse zero-sum statements less than younger people. They may also be less zero-sum because they perceive less resource scarcity. Recall that zero-sum beliefs arise when resources are scarce (Davidai & Tepper, 2023), and this is true even when they are perceived to be scarce but are not actually scarce (Meegan, 2010). We therefore considered whether older individuals perceived less resource scarcity and, as a result, endorsed less zero-sum beliefs. To do so, we measured the level of scarcity individuals were experiencing by utilizing the material scarcity subscale of the Perceived Scarcity Scale (DeSousa et al., 2020; e.g., “I have had to borrow money from family or friends to pay my bills”). Material scarcity is generally defined as not having enough material resources, but it can also extend to subjective perceptions of not having enough to successfully function in society. Details on scale items can be found in the Supplemental Materials.

Results

Young individuals ($M = 4.34$, $SD = 1.99$) judged the nonzero-sum situation as significantly more zero-sum than older individuals ($M = 3.69$, $SD = 2.11$), $t(782.99) = -4.43$, 95% CI [-.93, -.36], $p < .001$, Cohen's $d = -.32$, 95% CI [-.46, -.17] (Figure 5). This replicates the findings of Studies 1 and 2.

The age effect was again replicated with general zero-sum beliefs. Young individuals ($M = 4.33$, $SD = 1.53$) held significantly more general zero-sum beliefs than older individuals ($M = 3.37$, $SD = 1.54$), $t(785.99) = -8.77$, 95% CI [-1.17, -.74], $p < .001$, Cohen's $d = -.62$, 95% CI [-.77, -.48]. By contrast, there was no age difference in the immigration item by age ($M_{\text{Young}} = 5.50$, $SD_{\text{Young}} = 1.23$; $M_{\text{Old}} = 5.63$, $SD_{\text{Old}} = 1.21$), and for the government item, older participants ($M = 6.39$, $SD = .73$) actually endorsed the belief more than young individuals ($M = 6.20$, $SD = .89$), $t(758.74) = 3.28$, 95% CI [.08, .30], Holm–Bonferroni adjusted $p = .002$, Cohen's $d = .23$, 95% CI [.09, .37] (Figure 6).

As in Studies 1 and 2, the justifications participants provided for their ratings were coded (0 = nonzero-sum, 1 = zero-sum, 999 = nondiagnostic). Three research assistants coded the explanations, consistent with the instructions and procedure of Study 2. To assess the agreement between the three raters on their independent coding, we computed Light's kappa and found good and substantial agreement (Light's $\kappa = .71$, $z = 43$, $p < .001$). Excluding responses that were nondiagnostic, we performed a chi-square test to determine if the amount of zero-sum and nonzero-sum responses differed by age. Indeed, the proportion of participants who provided reasonings containing zero-sum beliefs was significantly different by age, $\chi^2(1, N = 599) = 27.74$, 95% CI [−.30, −.14], $p < .001$. Sixty-one percent of young participants' justifications for the nonzero-sum scenario expressed zero-sum thinking, whereas only 39% of old participants' responses did.

As in Studies 1 and 2, we examined the relationship between demographic scores and general zero-sum beliefs by conducting a moderation analysis and found no evidence of significant moderators. As in Study 2, we considered the relationship between age, executive functioning, and zero-sum beliefs. To do so, we conducted correlation analysis by age on general zero-sum beliefs and executive functioning and found no significant relationships (Table 2).

Age	Cognitive measure	t	df	p	r	95% CI
Young	TMT-A	−.68	392	.50	−.03	[−.13, .06]
	TMT-B	1.05	392	.30	.05	[−.05, .15]
	Word List recall	.65	392	.52	.03	[−.06, .13]
Old	TMT-A	2.14	392	.03	.11	[.01, .20]
	TMT-B	1.36	392	.18	.07	[−.03, .17]
	Word List recall	.44	392	.66	.02	[−.08, .12]

Similar to Study 2, we also examined the relationship between situation-specific zero-sum judgments and executive functioning by performing a median split on the Trail Making Task data and again found that young participants who took longer on TMT-B (M = 4.54, SD = 1.94; low performers) evaluated the nonzero-sum scenario as significantly more zero-sum than high performers on this task (M = 4.14, SD = 2.02), t(391.27) = −2.04, 95% CI [−.80, −.01], p < .05, Cohen's d = −.21, 95% CI [−.40, −.01] (Figure 7). In other words, as in Study 2, high performers on this cognitive measure rated the nonzero-sum scenario as significantly less zero-sum. Last, we examined if cognitive performance predicted our observed effect and found no significant evidence.

Next, we examined the impact of age on positive thinking and perceived resource scarcity. Older adults (M = 29.7, SD = 4.62) were significantly more positive thinkers than younger adults (M = 26, SD = 7.24), t(667.61) = 8.56, 95% CI [2.85, 4.55], p < .001, Cohen's d = .61, 95% CI [.47, .75]. Conversely, younger adults (M = 15.7, SD = 6.42) reported significantly greater levels of perceived material scarcity than older adults (M = 12.60, SD = 4.83), t(729.81) = −7.62, 95% CI [−3.88, −2.29], p < .001, Cohen's d = −.54, 95% CI [−.68, −.40]. Hence, older individuals held more positive beliefs and were less resource scarce.

Next, as in Studies 1 and 2, we conducted a linear regression to evaluate the extent to which situation-specific judgments, age (young or old), executive functioning, positive thinking, or perceived resource scarcity could predict general zero-sum belief scores. We again find a significant model, F(5, 782) = 56, p < .001, $R^2$ = 0.26. All the predictors were significant except for executive functioning: situation-specific judgments (β = 0.26, p < .001), age (β = 0.54, p < .001), positive thinking (β = −0.03, p = .003), perceived resource scarcity (β = 0.05, p < .001), and executive functioning (β = 0.02, p = .82). These findings highlight the role of positive thinking and perceived resource scarcity, in addition to age and situation-specific judgments, in shaping general zero-sum beliefs.

Given the similar nature of measures across Studies 1–3, below, we present three tables that compare general zero-sum beliefs as well as situation-specific judgment across studies (Tables 3–5).

Discussion

Study 3 replicated the finding that younger individuals evaluate a nonzero-sum situation as significantly more zero-sum than older individuals. Importantly, the general zero-sum beliefs of older individuals are significantly lower than younger individuals when it comes to zero-sum, but not with beliefs that are irrelevant to zero-sum thinking. We found that positive thinking and resource scarcity explain the relationship between age and zero-sum thinking. Last, we extend previous findings by discovering that positive thinking and perceived resource scarcity are also key contributors to differences in general zero-sum beliefs among young and old individuals.

Study 4: Age on a Continuum

Studies 1–3 compared groups of young and old adults. In Study 4, we evaluated zero-sum beliefs along the age continuum. Zero-sum beliefs may decline linearly with age, or they might follow an inverse U-shaped function. This will be consistent with related patterns of change over the age continuum such as expressed competition (Mayr et al., 2012). For instance, while some theories would suggest that competitive preferences decrease with age, Mayr et al. (2012) discovered an exception to this, termed the “feisty fifties,” in that competitive preferences increase and peak around age 50 before dropping. Therefore, if zero-sum thinking follows a similar pattern, one could expect an increase from young to middle age, followed by a reduction in zero-sum beliefs toward older adulthood.

Method

Participants

Based on a simple heuristic of 100 participants per age decade from 18 to 78, we aimed to recruit 600 people. Our final sample consisted of 601 participants who successfully completed an attention check, were native English speakers, and were aged between 18 and 78 years of age. A total of 304 (50.6%) identified as females, 446 (74%) identified as White, and 389 (64.7%) had at least a trade school certification. The same set of demographics as Study 1 was also examined (see the Supplemental Materials).

As in Studies 2 and 3, to measure executive functioning, participants completed the TMT and Word List Paradigm. Participants were excluded and replaced if they scored worse than 3 standard deviations from the mean in each task. To be included in this study, participants must have recalled at least 2.83 words in the Word Paradigm List, completed the TMT-A task in under 51.66 s, and completed the TMT-B task in under 137.19 s.

Materials and Procedure

We used the same materials and measures as Studies 2 and 3. To identify potential mechanisms, we again utilized positive thinking and perceived resource scarcity from Study 3 but added a measure of perceived time horizons. Given that perceived time left to live has been shown to exert a profound influence on motivation, we measured perceived time horizons by presenting three items from the Future Time Horizons scale (Carstensen & Lang, 1996) that load onto future time extension (Rohr et al., 2017; e.g., “My future seems infinite to me”). Details on scale items can be found in the Supplemental Materials.

Participants completed the same procedure as Studies 2 and 3 with minor differences. When evaluating the workplace scenario, participants responded to “Based on the evaluation system you just read about, to what extent do you believe that if one employee succeeds, another must fail? 1 = not at all; 7 = very much so.” After evaluating the scenario, participants completed an attention check, general zero-sum scale items, the P-scale, the Perceived Scarcity Scale, and the Future Time Horizons scale. Last, participants completed demographic and cognitive measures.

Results

The relationship between age and situation-specific zero-sum judgments was not linear. Thus, per our preregistration, we proceeded by grouping the data into three groups: young age (18- to ≤38-year-olds), middle age (<38- to ≤58-year-olds), and old age (<58- to 78-year-olds). We conducted a two-way ANOVA to examine situation-specific judgments and found a significant main effect of scenario type, $F(1, 595) = 77.22, p < .001, \eta^2 = .11, 95\%$ CI [.08, 1], and of age group, $F(2, 595) = 3.36, p = .04, \eta^2 = .01, 95\%$ CI [0, 1], on situation-specific judgments, but no significant interaction between the two ($p = .34$; Figure 8). Post hoc analysis of linear contrasts indicated that condition type significantly impacted zero-sum ratings ($t = 8.76, p < .001$). Linear contrast analysis for the age category was not statistically significant ($p = .28$). Overall, we found that in the zero-sum scenario, ratings reduced with age, with young individuals ($M = 5.40, SD = 1.90$) providing the highest ratings and old individuals providing the lowest ($M = 4.98, SD = 2.06$). We found a similar pattern in the nonzero-sum scenario, in that older individuals ($M = 3.42, SD = 2.10$) provided lower ratings than young individuals ($M = 3.73, SD = 1.95$), yet there was an inverted U-shape whereby ratings increased for middle-aged individuals ($M = 4.01, SD = 2.15$) before decreasing in old age.

General zero-sum beliefs significantly reduced by age, $F(2, 598) = 23.21, p < .001, \eta^2 = .07, 95\%$ CI [.04, 1] (Figure 9). Bonferroni post hoc tests revealed that the young age group had significantly higher general zero-sum beliefs ($M = 4.29$, $SD = 1.49$) compared to the middle age group ($M = 3.81$, $SD = 1.53$; $p = .006$) and to the old age group ($M = 3.23$, $SD = 1.63$; $p < .001$). Additionally, the old age group ($M = 3.23$, $SD = 1.63$) had significantly lower general zero-sum beliefs compared to the middle age group ($M = 3.81$, $SD = 1.53$; $p = .001$). Post hoc linear contrasts showed a significant linear effect for age category ($p < .001$), suggesting that as age increases, general zero-sum beliefs reduced significantly. Given the distribution of ages captured in the data, these results support that general zero-sum beliefs decrease with age.

Next, as in Studies 1–3, we conducted a linear regression to evaluate the extent to which situation-specific judgments (nonzero-sum), age (continuous), executive functioning, positive thinking, perceived resource scarcity, and the Future Time Horizons scale could predict general zero-sum belief scores. As with all previous studies, we find a significant model, $F(6, 290) = 22.7$, $p < .001$, $R^2 = 0.32$. Replicating previous studies, we find situation-specific judgments ($\beta = 0.36$, $p < .001$), positive thinking ($\beta = -0.03$, $p = .04$), and perceived resource scarcity ($\beta = 0.04$, $p < .001$) were significant predictors. Conversely, executive functioning ($\beta = 0.21$, $p = .08$), future time horizons ($\beta = 0.01$, $p = .86$), and age ($\beta = -0.01$, $p = .06$) were not. Given that splitting the data by scenario divides the sample in half, it is possible that a lack of power may explain why age did not significantly predict general zero-sum beliefs, although it was close ($p = .06$).

We also performed the same analysis we conducted in Studies 2 and 3 by splitting the data into young (20- to 35-year-olds) and old (65- to 80-year-olds) and examining the situation-specific zero-sum judgments. We split the data by scenario type and conducted $t$ tests. The differences between young and old groups were not significant for either scenario (nonzero-sum: $M_{\text{Young}} = 3.69$, $M_{\text{Old}} = 3.66$; zero-sum: $M_{\text{Young}} = 5.30$, $M_{\text{Old}} = 5.06$). We conducted a sensitivity analysis to examine whether the process of splitting the data resulted in enough power. A sensitivity analysis using G*Power3.1 (Faul et al., 2009) suggests an independent-samples $t$ test with 134 participants in total for the nonzero-sum scenario ($n_{\text{Young}} = 81$, $n_{\text{Old}} = 53$) and 151 participants in the zero-sum scenario ($n_{\text{Young}} = 99$, $n_{\text{Old}} = 52$) would be sensitive to reliably detect effects of Cohen’s $d = .50$ and Cohen’s $d = .48$, respectively, with 80% power ($\alpha = .05$, two-tailed). This suggests that the analysis is highly underpowered and is unlikely to detect a smaller effect size.

As in Studies 1–3, the justifications participants provided for their ratings were coded ($0 = \text{nonzero-sum}$, $1 = \text{zero-sum}$, $999 = \text{nondiagnostic}$). The same coding procedure from Studies 2 and 3 was employed. To assess the agreement between the three raters on their independent coding, we computed Light’s kappa and found good and substantial agreement (Light’s $\kappa = .70$, $z = 22.1$, $p < .001$). Excluding responses that were nondiagnostic, we performed a chi-square test to determine if the amount of zero-sum and nonzero-sum responses differed by age group (young, middle, old) and scenario type. For the nonzero-sum scenario, 50% of the young age group provided zero-sum justifications, whereas 47.6% of the middle age group and 31.4% of the old age group did, resulting in a significant chi-square, $\chi^2(2, N = 246) = 6.98$, $p = .03$. In the zero-sum scenario, participants across age groups provided similar rates of zero-sum justifications (young = 89%, middle = 79.5%, and old = 81.9%); these differences were not significant ($p = .19$). Taken together, while there were no significant differences in the situation-specific ratings, the justifications participants provided for their ratings suggest that indeed, situation-specific zero-sum judgments significantly reduce with age for situations that are not zero-sum.

Last, as in Studies 1–3, we examined whether education, gender, political beliefs, net worth, income, being retired, or having children moderated the relationship between age (continuous) and general zero-sum beliefs. We found no evidence of significant moderators.

Discussion

Extending our findings, we discover that zero-sum beliefs differ along the age continuum. While the relationship between age and zero-sum beliefs is not linear, by bracketing age into groups, young, middle, and old, we find a significant reduction in general zero-sum beliefs. Moreover, the reduction in zero-sum beliefs can be partly explained by positive thinking and material resource scarcity. With age, individuals become more positive thinkers and perceive less material resource scarcity, leading them to endorse less general zero-sum beliefs.

Study 5: Word Values Survey (WVS)

Our findings show an age effect whereby zero-sum beliefs differ with age. One might also consider an explanation for this effect that is unrelated to aging. That is, older individuals may exhibit less zero-sum thinking not because this thinking has decreased as they have aged throughout their lifespan; rather, these individuals, as a generation, are simply less zero-sum. In other words, it is possible that our older adults had already been less zero-sum in their thinking even when they were young. This would mean that what makes older individuals less zero-sum is being born and educated during a particular time period or being part of a generation. For example, being born in the 1950s and being exposed to the unique circumstances of that time might have made them less zero-sum compared to being born in the 2000s (regardless of age). While Experiments 1–4 do not allow us to disentangle these two ideas, the data afforded by the WVS research program do.

Study	General Zero-Sum Belief No. 1 (M)	General Zero-Sum Belief No. 2 (M)	General Zero-Sum Belief No. 3 (M)	Average General Zero-Sum (M)	Zero-Sum Situation-Specific (M)	Nonzero-Sum Situation-Specific (M)
Study 1
Young	4.71	4.21	5.10	4.39	3.85
Old	3.53	3.10	3.88	3.52	3.53
Study 2
Young	4.59	4.28	4.95	4.60	4.50	5.89
Old	3.54	3.13	3.85	3.50	3.71	5.68
Study 3
Young	4.38	3.93	4.66	4.33	4.34
Old	3.41	2.98	3.71	3.37	3.69

The goal of the WVS is to capture changes in social, political, and economic values over time. They do so by surveying tens of thousands of individuals every 5 years in 100+ countries all around the world. For example, countries can range from Angola, Bangladesh, and Haiti to Switzerland, the United States, and China. For information about the survey, see https://www.worldvaluessurvey.org/. Thus far, they have conducted seven waves of inquiry, resulting in a databank of hundreds of thousands of people. Importantly, they have a question relating to zero-sum beliefs whereby they ask respondents to indicate whether they believe that people can only get rich at the expense of others, as opposed to the idea that wealth can grow so there is enough for everyone. While no one individual is followed throughout their lifetime as part of this survey, data across decades allow us to compare generations and age groups to examine changes in zero-sum beliefs.

Specifically, we explore patterns in the responses of participants to the WVS in Waves 3 (1994–1999) and 6 (2012–2016). Having respondents from all over the globe allows us to look for age differences in zero-sum beliefs while holding respondents' country (and therefore many of their life experiences) constant. If our experimental findings are real and applicable to everyday people, then we should find that zero-sum beliefs decrease with age when using this data set. Moreover, because data were collected both in 1994–1999 (Wave 3) and in 2010–2016 (Wave 6), we can disentangle the effects of age from generation and evaluate if being part of a generation as indexed by year of birth impacts zero-sum beliefs and if age per se also affects them. This approach allows us to explore to what degree zero-sum beliefs are shaped by developmental changes associated with aging, by generational differences rooted in shared life experiences, and by the unique contextual factors present during the time of data collection. Specifically, the inclusion of data from Wave 3 (1994–1999) and Wave 6 (2010–2016) provides a unique opportunity to compare individuals of similar ages at two distinct points in time, as well as individuals from the same birth cohorts across different ages. By leveraging these two waves, we can disentangle the effects of age, generation, and historical context, providing a more comprehensive understanding of how zero-sum beliefs differ over time.

Method

Participants

Across Waves 3 and 6, 245,201 participants answered a zero-sum belief question. For the analysis, we excluded participants who had missing values for the demographic variables. For example, we excluded 114 participants who did not report their gender out of the two binary options offered. Additional demographic variables included missing values, so that the final sample we use in Table 6 (Model 2) is a sample of 207,071 participants that includes responses to all of the variables of interest (Table 6, Model 2). In other specifications (Table 6, Model 1, and Figure 10), we use larger samples that include all available data in the original data set. Ns are reported for each of these analyses. In Table 7, we present the descriptive statistics for the final data set, by wave.

The variable "zero-sum beliefs" captures participants' zero-sum beliefs on a scale of 1–10 (where 10 indicates expressing the strongest zero-sum beliefs; see our description below). The variables "age" and "year born" are continuous variables capturing participants' age and the year in which they were born. The variable “female” is a binary variable capturing participants reporting being female (approximately 51% of the entire sample). The variable “primary education” is a binary variable capturing reporting having primary education (approximately 12.7% of the entire sample), and the variable “secondary education” is a binary variable denoting reporting having secondary education (about 15.9% of the sample). The variable “more than secondary education” captures reporting having more than a secondary education (approximately 25% of the sample). Finally, the variable “socioeconomic class” is a continuous variable (on a scale of 1–5) capturing participants’ reported social class.

Study	Judgment	Test statistic	p	Effect size
Study 1	General zero-sum	$t(656.8) = -11.37$	<.001	Cohen's $d = -.88$
	Situation specific (nonzero-sum)	$t(664.18) = -2.03$	<.05	Cohen's $d = -.16$
Study 2	General zero-sum	$t(406.85) = -7.39$	<.001	Cohen's $d = -.72$
	Situation specific (nonzero-sum)	$t(206.98) = -2.76$	<.006	Cohen's $d = -.38$
Study 3	General zero-sum	$t(785.99) = -8.77$	<.001	Cohen's $d = -.62$
	Situation specific (nonzero-sum)	$t(782.99) = -4.43$	<.001	Cohen's $d = -.32$

Materials and Procedure

We focused on the “wealth accumulation” variable in the WVS that captures zero-sum thinking (Ongis & Davidai, 2022). This variable first appears in Wave 2 and last appears in Wave 6. Inconsistencies in the variable coding for Wave 2 prompted us to use Wave 3 instead of 2. In Waves 3 (Inglehart et al., 2014a) and 6 (Inglehart et al., 2014b), participants used a 10-point scale to indicate their belief: 1 = people can only get rich at the expense of others and 10 = wealth can grow so there is enough for everyone. They marked the number anywhere between 1 and 10 that corresponded to their belief on this continuum. To be consistent with the scales used in Studies 1–4, we reverse-coded these answers so that 10 represented the strongest zero-sum thinking.

Results

Using a subsample of participants from the two waves who were older than 20 and younger than 80 when the survey was conducted, we calculated the means of participants’ zero-sum beliefs by age group (Figure 10).

We see that young adults, ages 18 to 37 ($N = 93,165$), expressed significantly greater zero-sum beliefs compared to middle-aged and older adults, ages 38–57 ($N = 74,649$) and 58–77 ($N = 36,709; p < .01$). The average zero-sum beliefs for young adults was 4.71 ($SD = 2.81, 95\% \text{ CI } [4.69, 4.73]$) compared to 4.58 ($SD = 2.81, 95\% \text{ CI } [4.57, 4.61]$) for the middle-aged group and 4.59 ($SD = 2.81, 95\% \text{ CI } [4.56, 4.62]$) for the older adults.

To better understand the relationships among age and zero-sum beliefs, Figure 11 displays the standardized zero-sum beliefs of participants (on the y-axes), plotted by age (on the x-axes) and wave. This enables us to observe that by standardizing zero-sum beliefs across both waves, there is a sharp decrease by age. By standardizing zero-sum scores and plotting the two waves, we can assess that independent of wave, there is a decrease in zero-sum beliefs.

In Table 6, we present the results of ordinary least squares regression models predicting zero-sum beliefs using respondents’ age and birth year as key predictors. The models are progressively adjusted as follows:

Model 1 includes only age and year of birth as predictors.

Model 2 adds controls for demographic characteristics.

Model 3 introduces a control for survey wave, using a binary variable (“Wave 6”) to indicate respondents surveyed during Wave 6 (administered from 2010 to 2014, compared to Wave 3, administered from 1995 to 1999).

Model 4 replaces the continuous variable for birth year with a series of dummy variables, each representing a specific year of birth for the respondents.

All models incorporate country fixed effects to account for variation across respondents’ countries of residence. The primary goal of the analysis is to disentangle the effects of age from the effects of cohort (birth year). However, it is important to note that age, survey year, and birth year are perfectly linearly related because the sum of birth year and age equals the calendar year. This creates an identification problem if all three variables are treated as linear (Yang & Land, 2013). Our design overcomes this issue in the following ways: First, since each wave spans multiple years (e.g., Wave 3 from 1995 to 1999 and Wave 6 from 2010 to 2014), the wave variable is not perfectly correlated with the birth year. Each wave includes respondents from multiple birth years, providing variability across survey periods. This overlap allows us to separate the effects of the survey wave from the effects of cohort. Second, by including the wave dummy variable (e.g., 'Wave 6'), we account for contextual influences tied to the survey period. Third, by modeling cohort as a set of dummy variables for each birth year (Model 4), we can isolate cohort effects from age and wave effects.

Variable	Study 1	Study 2	Study 3
Model statistics	$F(2, 665) = 77.23, p < .001, R^2 = 0.19$	$F(3, 205) = 35.11, p < .001, R^2 = 0.34$	$F(5, 782) = 56, p < .001, R^2 = 0.26$
Situation-specific judgments	$\beta = 0.09, p < .001$	$\beta = 0.33, p < .001$	$\beta = 0.26, p < .001$
Age (young vs. old)	$\beta = 0.85, p < .001$	$\beta = 1.01, p < .001$	$\beta = 0.54, p < .001$
Executive functioning		$\beta = 0.06, p = .63$	$\beta = 0.02, p = .82$
Positive thinking			$\beta = -0.03, p = .003$
Resource scarcity			$\beta = 0.05, p < .001$

Variable	Model 1	Model 2	Model 3	Model 4
Ordinary least squares regression models predicting zero-sum beliefs (country fixed effects)
Age	0.018*** (0.001)	0.017*** (0.001)	-0.028** (0.013)	-0.026** (0.013)
Year born	0.022*** (0.001)	0.022*** (0.001)	-0.023 (0.013)
Wave 6			0.243*** (0.069)	0.243*** (0.069)
Female		-0.092*** (0.012)	-0.092*** (0.012)	-0.092*** (0.012)
Socioeconomic class		0.164*** (0.007)	0.165*** (0.007)	0.165*** (0.007)
Primary education		-0.025 (0.020)	-0.025 (0.020)	-0.025 (0.020)
Secondary education		-0.037** (0.018)	-0.037** (0.018)	-0.037** (0.018)
More than secondary education		-0.047*** (0.016)	-0.047*** (0.016)	-0.047*** (0.016)
Constant		-40.476*** (1.997)	49.380 (25.503)	6.500*** (1.691)
Country (dummies)	+	+	+	+
Year born (dummies)				+
N	214,197	207,071	207,071	207,071

Variable	Wave 3	Wave 6	Minimum value	Maximum value
	M/SD	M/SD
Zero-sum beliefs	4.5632.827	4.7462.775	1	10
Age	41.10715.845	41.68916.401	15	102
Female	0.5140.500	0.5180.500
Year born	1955.03715.847	1970.42116.587	1,901	1,998
Primary education	0.1340.341	0.1080.310
Secondary education	0.1550.362	0.1760.380
More than secondary education	0.2570.437	0.2620.440
Socioeconomic class	3.3100.940	3.3030.998	1	5
Observations	124,346	82,725

Thus, only Models 3 and 4 effectively disentangle age from cohort effects while also holding constant the influence of the survey wave and real-world events that may have occurred during specific survey periods. Note that dummy-coding birth years (Model 4), rather than treating this variable as continuous (Model 3), allows us to capture nonlinear cohort effects that a continuous variable would miss. By treating birth year as a set of discrete categories, we can account for potential idiosyncratic or cohort-specific events that may have uniquely influenced individuals born in specific time periods. For example, significant historical, cultural, or economic events occurring during a cohort's formative years—or even in their parents' lives—could have long-lasting impacts on attitudes, such as zero-sum beliefs. These kinds of effects are unlikely to be adequately represented by a linear trend.

Moreover, dummy-coding relaxes the assumption that the relationship between birth year and zero-sum beliefs is continuous or linear. This is particularly relevant when analyzing data across multiple decades, as the impact of being born in, say, 1940 may not simply differ incrementally from being born in 1939 or 1941, but may reflect unique influences associated with a specific historical context.

Importantly, because age, birth year, and survey year are mathematically interrelated (age + birth year $\approx$ survey year), including all three in the same model introduces multicollinearity, which inflates standard errors. This issue is evident in Model 3, where variance inflation factors (VIFs) are extremely high. In Model 4, we address this by dummy-coding birth year to allow for more flexible cohort-specific effects. While this approach does not eliminate multicollinearity, it offers an alternative specification that makes fewer assumptions about the functional form of the relationship.

Taken together, the results in all models suggest that age is significantly and negatively correlated with holding greater zero-sum beliefs. Interestingly, when the wave (the Wave 6 dummy) is not controlled for, the effects of age in both Model 1 and Model 2 are significant and positive. Nonetheless, because we wish to disentangle the effects of age, birth year, and survey timing, controlling for the wave (in both Model 3 and Model 4) provides a more accurate analysis.1 Indeed, we find that controlling for the survey timing (wave) and cohort (birth year as a continuous variable and as a set of dummies) generates a significantly negative effect for age.

In Model 3, being 1 year older when asked the question resulted in a 0.028 decrease in the zero-sum beliefs expressed ($p < .01$). Interestingly, in this model, being born a year later was positively correlated with a 0.023 increase in one's zero-sum beliefs ($p < .01$). Recall that because the data were collected in different years, we can disentangle the effects of age from the effects of the year in which one was born. To illustrate the results of Model 3, we plot them in Figure 12.

In sum, the analysis allows us to disentangle the effects of age from the effects of cohort while holding constant the real-world events at the time of participating in the survey (using the Wave 6 dummy and the country dummies). The findings suggest that being older is negatively and significantly correlated with zero-sum beliefs.

Discussion

Data from around the world show that general zero-sum beliefs decrease with age for two reasons. First, being older reduces people's zero-sum beliefs. Second, being born earlier also reduces zero-sum beliefs, meaning that today’s older generation held fewer zero-sum beliefs even when its members were young. This demonstrates that both age and generation contributed to our experimental findings that older adults exhibit less zero-sum thinking. General zero-sum beliefs decrease with age independent of how zero-sum a generation is, highlighting the significance of the age effect in this bias.

General Discussion

We discovered that people’s zero-sum beliefs differ with age. Zero-sum beliefs can be operationalized in terms of general beliefs that influence how people perceive the world around them or in terms of situation-specific judgments of specific situations. Across studies, we consistently find that younger individuals hold significantly more general zero-sum beliefs (Studies 1–4) and that their judgment of specific situations is also more zero-sum (Studies 1, 2, and 4). We consistently find that this relates to older people becoming more positive thinkers with age and experiencing less resource scarcity (Studies 3 and 4). To evaluate whether a reduction in zero-sum thinking is the result of being older or being of a different generation, we analyzed WVS data from a quarter of a million people around the world (Study 5). By using these survey data, we show that a decrease in zero-sum beliefs by age replicates within countries around the world, and we are able to tease apart a true age effect from generational differences. We demonstrate that both age and generation differences independently contribute to the effects we find.

Our findings that perceived resource scarcity and positive thinking partly explain why older individuals have lower zero-sum beliefs extend past research showing an association between resource scarcity and greater zero-sum beliefs (Meegan, 2010) and that older individuals are more positive thinkers (Carstensen & DeLiema, 2018). Rather than having individuals compare themselves to others (Gerber et al., 2018; Ongis & Davidai, 2022), we find that one's inherent experience about their personal material scarcity is related to the degree to which they endorse zero-sum beliefs. Similar to positive thinking shaping how older people attend and remember information (Mather & Carstensen, 2005), our research suggests that positive thinking might also safeguard against biases. Indeed, positive thinking led older adults to evaluate nonzero-sum situation information in a less zero-sum manner compared to younger individuals. Thus, our research suggests that older people's enhanced positive thinking may mitigate zero-sum beliefs.

Broadly, this research ties well with findings that demonstrate changes in beliefs and behaviors with age. For example, older adults are more trusting (Bailey & Leon, 2019), more prosocial (Beadle et al., 2015; Mayr & Freund, 2020), and more charitable (Bekkers & Wiepking, 2011) than younger individuals. For example, older participants are more likely to trust others in a trust game compared to younger participants even when the recipient has a reputation for being untrustworthy (Bailey et al., 2016). Moreover, a working article by Khon et al. (2021) finds that age is negatively associated with the belief that marketers could manipulate people. Related to zero-sum thinking, the authors conclude that younger individuals should be encouraged to think about the customer and the marketer's side to help attenuate beliefs. Therefore, our finding that zero-sum thinking decreases with age could perhaps explain some findings in the marketing and decision-making literature and beyond.

Constraints on Generality

Our experiments recruited individuals across the age continuum from 18 to 80 years old to address how age may impact zero-sum thinking. We specifically used online populations, which allowed us to reach participants from different demographic backgrounds and locations. While older individuals that are on Prolific may be more tech savvy than the general population, utilizing Prolific allowed us to test samples that might be more inclusive. For instance, the same type of selection bias would be true for in-person research as it would be inclusive of older individuals who are fit enough to travel and participate in the in-person study. We found the best way to address this concern was by including the WVS as this collects data across countries, in urban and rural populations. Notwithstanding, while the WVS attempts to address the generalizability of our discovery and replicates the patterns found in our experimental data, we have no longitudinal evidence. Thus, future research would benefit from a longitudinal approach whereby people's zero-sum beliefs are tracked for decades throughout their lifetime.

The type of scenarios utilized in this body of research relates to the workplace. It is possible that individuals may have construed the situation in different ways (distal vs. proximal) depending on age or experience. To account for this, we collected employment information across studies. For the young cohort, most reported either working part-time or full-time (61.3% for Study 1, 65% for Study 2, and 70% for Study 4); thus, while we do not know how much time they have been in the workforce, they are certainly in the workforce. Yet, additional research using other contexts across ages is warranted for broader generalizability.

Practical Implications

Our results are of practical importance as zero-sum beliefs extend to everyday activities like negotiation. Zero-sum thinking may lead people to a fixed-pie bias (de Dreu et al., 2000). This often results in not inquiring about the other person's interests and failing to consider how one could create value through win-win integration. While there is little research as to the age differences that impact negotiation outcomes (Boothby et al., 2023; Elfenbein, 2014), our research suggests that older people may be more integrative than younger individuals since they are less zero-sum. By having young and old people negotiate with and against each other, Kappes et al. (2020) found the opposite: that younger individuals were more integrative than older individuals. The success of younger individuals can be explained by information exchange, as younger individuals exchanged more information about their priorities than older individuals. It is unclear how zero-sum beliefs would affect negotiation in these studies as the researchers did not measure zero-sum thinking. It is possible, then, that one's zero-sum beliefs still impact how one negotiates or behaves in different contexts. It would be worth investigating the role zero-sum beliefs might play in negotiation and whether it leads to behavioral differences among young and old individuals in that context.

The current research illustrates how zero-sum thinking differs with age. One may ask, why does this matter? Populism in politics across the political spectrum tends to promote claims that resources are zero-sum, mostly as a tool to gain votes (Ali et al., 2025; Papaioannou et al., 2023). We have seen this to be true in politics—Senator Bernie Sanders's message during his 2016 and 2020 presidential runs was that the wealthy were getting rich at the expense of the poor. His campaign website explicitly highlights this belief, "there has been a massive transfer of wealth from those who have too little to those who have too much" (Bernie Sanders, n.d.). If elected officials are communicating zero-sum beliefs, one must consider the practical implications it will have on people's behavior. Our research suggests that younger people might be more receptive to these ideas, as they are more likely to endorse zero-sum beliefs than older individuals. In a world where people are living longer and one in six people is expected to be over 60 by 2030 (World Health Organization, 2022), we must seriously consider these findings.

Conclusion

Zero-sum beliefs are pervasive and impact how people understand the world and behave in it. And yet, we discover that these beliefs differ with age. Specifically, older individuals endorse less zero-sum thinking than younger individuals. This is the case around the world, and it might result from both aging and generational differences. We cannot predict how zero-sum future generations will be in their thinking. The good news is that whatever their generational baseline is, growing older would make them see the world as less zero-sum. This will surely have its benefits.

Footnotes (1)

1 VIFs for all independent variables were examined. Most variables had VIFs between 1 and 5. The variables for age and year of birth had VIFs between 5 and 10 in Models 1 and 2. As expected, in Models 3 and 4—where both age and birth year were included along with survey wave—the VIFs increased substantially, reaching values above 1,500 due to collinearity (age + birth year $\approx$ survey year). While dummy-coding birth year in Model 4 does not fully resolve the multicollinearity, it provides an alternative specification that allows for more flexible interpretation of cohort effects.