Study with the several resources on Docsity
Earn points by helping other students or get them with a premium plan
Prepare for your exams
Study with the several resources on Docsity
Earn points to download
Earn points by helping other students or get them with a premium plan
Community
Ask the community for help and clear up your study doubts
Discover the best universities in your country according to Docsity users
Free resources
Download our free guides on studying techniques, anxiety management strategies, and thesis advice from Docsity tutors
The relationship between collectivism, perceived group diversity, and group performance on conjunctive and disjunctive tasks. The study finds that collectivism benefits group performance on conjunctive tasks with high objective national diversity, but harms group performance on disjunctive tasks with high objective expertness diversity. The research also identifies the role of task type in understanding how collectivism influences perceived group diversity and its impact on group performance.
Typology: Study notes
1 / 49
This page cannot be seen from the preview
Don't miss anything!
Blurred Lines: How the Collectivism Norm Operates Through Group Perceived Diversity to Boost or Harm Group Performance Jennifer A. Chatman University of California, Berkeley Lindreed Greer Stanford University Eliot Sherman London Business School Bernadette Doerr Forthcoming, Organization Science October 9, 2018
Abstract We develop and test a theory that reconciles contradictions in how collectivistic norms influence group performance. We draw on the perceived diversity literature to hypothesize that collectivistic norms cause group members to “blur” demographic differences, resulting in a shared perception that group members are more similar to one another than they actually are. Whether this benefits or harms group performance depends on the level of objective diversity in the group and the relevance of the perceived diversity attribute for accomplishing the group’s task. For conjunctive tasks, the group’s performance is determined by its weakest member, and high levels of cohesion are needed. Our theory suggests that collectivism benefits group conjunctive performance when objective national diversity is high by blurring divisive relational differences but has no effect in groups with low objective national diversity. In contrast, for disjunctive tasks, the group’s performance is determined by its best member, and we predict that collectivism harms group disjunctive performance when objective expertness diversity is high by blurring differences in task-relevant expertness, but has no effect in groups with low objective expertness diversity. We find support for our theory in two studies, including an archival study of 5,214 Himalayan climbing groups and a laboratory experiment assessing 356 groups. Our results show that collectivism has benefits and detriments for diverse groups, and that these contradictory effects can be understood by identifying how the collectivistic blurring of perceived group diversity helps or hurts groups based on the type of tasks on which they are working.
component of collectivism and, we argue, can usefully be examined as a central mechanism in how collectivism operates in groups. We identify how, in bringing group members together, collectivism has the potential to help or hinder group performance because of how it influences perceived group diversity, defined as “the degree to which members are aware of one another’s differences, as reflected in their internal mental representations of the unit’s composition” (Shemla et al. 2016). By making salient a view of members as interchangeable exemplars of the group (Chatman et al. 1998), the norm to be collectivistic reduces group members’ perceived diversity—their propensity to view one another as different—and effectively causes them to “blur” distinctions between them, much like an out-of-focus group photo that makes it difficult to distinguish between individuals. A key implication of blurring is that perceptions of diversity may not correspond to objective measures of diversity as closely as was once thought (Pfeffer 1983); and that, accordingly, perceived diversity may exert a greater influence on group processes and outcomes than does a group’s objective diversity (Shemla et al. 2016). To understand which group contexts are most influenced either positively or negatively by the collectivistic blurring of a group’s diversity, we build upon Bell’s (2007) suggestion to examine the intersection between tasks—whether they are more conjunctive or disjunctive (LePine et al. 1997, Steiner 1972)—and diversity type—in which we differentiate between relations- and task-oriented diversity (Jackson et al. 1995). Specifically, we propose that, to be successful on conjunctive group tasks (tasks on which all members of the group must complete the task and the group cannot proceed to the next task until each member has done so; Frank and Anderson 1971), groups depend upon a strong sense of solidarity among members (Goncalo et al. 2010, Miller and Komorita 1995). In such settings, cohesion is essential, and perceiving relations-based differences among group members (such as national differences, which are typically less task-relevant) may particularly impede the solidarity needed to perform well (Williams and O’Reilly 1998). Thus, groups working on conjunctive tasks with a high level of objective relations-based diversity may benefit the most from a collectivistic norm, which reduces perceptions of relations-based diversity and enables group solidarity. In contrast, we propose that group tasks that are disjunctive in nature (where performance is a function of the most competent members of the group—and in which, therefore, individuation and
expertise-identification are critical) are most helped by maintaining salient differences among group members, particularly those related to task-oriented diversity, or the distribution of attributes that are relevant to a group’s work such as expertness (Bunderson and Sutcliffe 2002). Thus, groups working on disjunctive tasks with a high level of objective task-based diversity may be the most harmed by a collectivistic norm, which reduces perceptions of task-based diversity and thereby blocks a group’s ability to identify members’ expertness levels (Littlepage 1991). We conducted two studies to test our theoretical model. Using an archival study of Himalayan mountain-climbing expeditions, we demonstrate that collectivism benefits group performance on a conjunctive task (summiting a mountain, as discussed by Steiner, 1972), and harms group performance on a disjunctive task (recognizing differences in members’ expertness to preserve climber safety). We chose the Himalayan expedition context because the different sub-tasks incorporated within it include both conjunctive group tasks, which require cohesion and pro-social helping, as well as disjunctive tasks, which necessitate complex problem solving. Additionally, the outcomes in this setting are objective and immensely consequential, which is often not the case for empirical tests of group processes (Van Dijk et al. 2012). In our second study, a controlled laboratory experiment with 366 three-person groups, we investigate the mechanism explaining our effects shown in Study 1 and show in a causal test of our model that collectivism causes members to blur their perceptions of all forms of group diversity, explaining how the norm operates to benefit or harm group performance. Across both studies, we show that collectivism benefits group conjunctive task performance in groups with high objective levels of nationality diversity and harms group disjunctive task performance in groups with high objective levels of expertness diversity. We offer three contributions in this paper. First, we address the ongoing debate about the effectiveness of a collectivistic orientation for group outcomes (Goncalo and Staw 2006). We show that the contradictory effects of collectivism on group performance can be reconciled by considering both group composition (the levels of relations- and task-oriented diversity) and the type of task (whether group performance is dependent on group cohesion (conjunctive tasks) or on individual differentiation (disjunctive tasks)). Second, we offer insight into the primary mechanism by which collectivism benefits
effects on groups. We propose that the key mechanism by which collectivism influences groups is by reducing the perception of group diversity, which may help in certain contexts and hurt in others. Collectivism and Perceived Group Diversity Collectivism at the task-group level is best understood as a social norm (Jetten et al. 2002), or a widely held belief regarding the appropriateness of behaviors such as a concern for maintaining harmony and commitment to shared goals within the group (Cialdini et al. 1990). People follow social norms closely to verify that their behavior and beliefs are appropriately aligned with relevant groups’ expectations (Fu et al. 2007). Thus, greater exposure to the norm to be collectivistic should orient group members toward shared goals and predispose them to work towards the group’s collective interests (Marcus and Le 2013). Such factors can lead members to see the group as a viable social category with which to identify (Ashforth and Mael, 1989) and, correspondingly, to begin to focus less on each other’s individuating attributes and more on their shared similarities. Such tendencies toward depersonalization (Turner, 1982) lead members to see themselves as relatively interchangeable exemplars of the group. To understand how collectivism depersonalizes members and influences groups, we draw on theories of group diversity (Joshi and Roh 2009; Van Knippenberg et al. 2004). Researchers have been keenly interested in group diversity for over 60 years, but the relationship between objective diversity and group performance remains decidedly ambiguous, with multiple meta-analyses and reviews finding no, or only small and often inconsistent direct effects (Mannix and Neale 2005, Williams and O’Reilly 1998). To address these mixed findings, scholars initially attempted to distinguish between different forms of diversity. For example, Jackson and her colleagues (1995) distinguished between “relations-oriented diversity” and “task-oriented diversity.” Relations-oriented diversity, which includes differences in nationality or gender, is relevant to how interpersonal relationships evolve within a work group. When these attributes are made salient, people in stereotyped categories such as sex and race are perceived as less effective and prevented from making contributions regardless of their actual qualifications (Fiske and Markus 2012). Such biases and categorizations can harm social cohesion (Jehn et al. 1999), hindering overall group performance (Hornsey and Hogg 2000). In contrast, task-oriented diversity, such as differences in tenure and functional background (Van der Vegt et al. 2005), reflects the distribution of
attributes that are relevant to a group’s work. These attributes constitute a group’s cognitive resource base and can be associated with elaboration-based processes such as information exchange, information processing, gaining feedback, and integrating information (Joshi and Roh 2009). Some research shows that these differences in education, tenure, and function increase group performance (Jehn et al. 1999) and group creativity (Bell et al. 2011). Critically, however, meta-analyses on the effects of objective diversity on group performance have failed to find robust differences between relations- and task-oriented diversity (Horwitz and Horwitz 2007, Van Dijk et al. 2012). Recent research suggests that the lack of a consistent relationship between objective diversity and group performance may be due to the fact that diversity is not always perceived in groups (Shemla et al. 2016). Scholars have long observed that objective differences are not always recognized by group members (Zellmer-Bruhn et al. 2008) and do not necessarily reflect differences in perspectives or informational resources (Lawrence 1997). Indeed, the most prominent theories of how diversity influences groups are predicated on members’ acknowledging differences—as an antecedent of group processes such as information elaboration (Van Knippenberg et al. 2004) and conflict (Allport 1954). Despite the evidence for the decoupling between objective and perceived diversity, diversity research has predominantly relied on “black-box” perspectives which account only for the actuarial presence of diversity and has failed to find consistent links to group processes and performance (Shemla and Meyer 2012). Therefore, it is increasingly apparent that researchers need to understand when and how objective diversity is perceived in groups (Van Dijk et al. 2012), particularly given the increases in diversity in the workplace (Mannix and Neale 2005). The emerging work on perceptions of group diversity builds on social identity theory, which suggests that group members make automatic judgments about the objective level of diversity in their group (Phillips et al. 2018). Initial research suggests that a focus on perceived group diversity may yield more robust theories and consonant findings than have approaches based on objective measures of diversity (Shemla et al. 2016). For example, objective demographic faultlines harmed group outcomes when group diversity perceptions were high, but not when they were low (Homan and Greer 2013, Jehn and Bezrukova 2010). While this line of work is encouraging, one of the key challenges is that
In understanding when collectivism causes members to blur different forms of perceived group diversity and how blurring, in turn, influences group performance, we look to research on the types of tasks on which groups are working. Researchers have shown that group tasks vary significantly in how much and what type of interdependence is needed to achieve success (McGrath et al. 1995, Steiner 1972). A key task-distinction that maps well onto the categorization versus elaboration processes that characterize groups that perceive themselves to be relationally- or task-diverse (Van Knippenberg et al.
task performance may be particularly bolstered when members perceive task-oriented differences that promote the individuation and information elaboration needed to identify expert members and give them greater influence in group decisions (Van Knippenberg et al. 2004). Therefore, we suggest that the collectivism norm can benefit group performance on conjunctive tasks by reducing members’ perception of relations-oriented diversity, but that it can also harm group performance on disjunctive tasks by reducing members’ perception of task-oriented diversity. Collectivism improves conjunctive task performance in objectively relationally diverse groups. We first propose that emphasizing the collectivism norm will improve conjunctive task performance, but only in groups that have high, but not low, objective relations-oriented diversity. We focus here on objective national diversity, a specific and common relations-oriented diversity attribute. Nationality is a fundamental aspect of a person’s identity because it is in force “from birth,” is readily detectable, and influences how people interact, share information, and define and solve problems (Gibson and Gibbs 2006). Cultural identity groups share certain worldviews, sociocultural heritages, norms, and values (Pieterse et al. 2013). As such, the perception of different nationalities within task groups can be disruptive. When national differences are actively perceived, members are sensitized to differences in eye contact, punctuality, physiological reactions to emotional stimuli, and conversational style, and disagreements about the appropriateness of these behaviors can produce affective conflict and reduce group performance (De Dreu and Weingart 2003). We suggest that in objectively nationally diverse work groups, collectivism reduces the perception of national diversity, and thereby improves group performance on conjunctive tasks. For tasks requiring a conjunctive orientation, a significant risk is not establishing the cohesion and group motivation needed to ensure that all members are successful and can advance to accomplish the group’s goals (Weber and Hertel 2007). Collectivism can be critical for groups with high objective national diversity by blurring the perceptual categories associated with national diversity—perceptual categories the most likely to stand in the way of the needed group cohesion. This is because collectivism can help groups by redirecting perceptual attention from individual social category differences and toward shared common group attributes which can boost shared norm adherence, cooperation, commitment, group cohesion, and
are more cooperative than groups that lack either diversity or collectivism (Flynn et al. 2001) and that they perform better (Chatman et al. 1998, Goncalo and Staw 2006). The positive association between collectivism and cooperation in the relationally-diverse groups studied by Wagner (1995) and Eby and Dobbins (1997) is consistent with this as well. Collectivism harms disjunctive task performance in objectively task-diverse groups. In contrast to the positive impact that the collectivistic blurring of perceived relations-oriented diversity has on conjunctive task performance, we propose that collectivism will harm disjunctive task performance in groups characterized by more (but not less) objective task diversity. We focus here on objective expertness diversity, which reflects the degree to which members have different levels of task-relevant knowledge (Bunderson and Sutcliffe 2002), and is distinct from expertise diversity, which focuses on the type of expertise (similar to functional differences) (Van der Vegt et al. 2006). The perception of objective expertness diversity is particularly relevant as disjunctive task performance is determined by identifying the single best solution on behalf of the entire group (Steiner 1972). Identifying and choosing a single solution to a complex problem on behalf of a group necessitates an expertness hierarchy within the group, which can provide a heuristic for weighting information and identifying whom should make the final decision on behalf of the group (Keltner et al. 2008). We predict that collectivism will harm disjunctive task performance in more objectively expertness diverse groups by blurring this potentially valuable form of diversity. This is because by blurring the perception of differences that are relevant to accomplishing disjunctive tasks, collectivism diminishes the elaboration capacity of objective expertness diversity (Gardner 2012). Although having some members with more expertness on a complex task has been argued to provide a useful hierarchy to coordinate information and facilitate group performance (Greer et al. 2018), benefiting from expertness diversity depends on a group’s ability to accurately recognize, incorporate, and weight it (Pieterse et al. 2013). According to the categorization-elaboration model, one of the key antecedents of knowledge elaboration is the belief that other group members are sources of novel and useful information and perspectives (Van Knippenberg et al. 2004). Furthermore, demography research in the information/decision-making tradition suggests that perceiving task-oriented diversity, such as expertness, enables groups to air and
parse conflicting viewpoints, thereby improving their performance (De Dreu and West 2001). Thus, collectivism is likely to hurt disjunctive task performance in objectively task diverse groups by blurring the perceived task-oriented diversity needed to foster the dissent and task conflict required to surface relevant information (Thompson et al. 1996), and weight member’s perspectives in accordance with their task knowledge (Molden and Higgins 2005, Wang and Lee 2006). This appears inconsistent with research supporting the benefits of collectivism on group performance (Bell 2007), but it is consistent with the idea that success on creative, complex tasks requires the group to identify divergent knowledge and capitalize upon it—precisely the kind of activity that blurring obstructs, insofar as group members’ subjective perceptions of similarity preclude them from surfacing the unique information that other members actually possess (Phillips et al. 2004). The blurring of such task differences in groups that are collectivistic and diverse in terms of objective expertness can thereby foster even lower levels of information elaboration and expertness identification than is seen in either homogeneous groups (with low or high levels of collectivism) or objectively expertness diverse groups with low levels of collectivism which do perceive their differences. This occurs because, while stronger collectivism norms are expected to mute (and not completely obscure) objective differences in members’ expertness, there nonetheless is likely to be at least some awareness of these differences. When it comes to expertness, because it is a source of respect and status in groups (e.g., Bunderson & Sutcliffe 2002), members may want such differences (even if perceived as modest) to be acknowledged in some way. Yet this may be less likely to happen in such groups. Rather, perceptions of similarity may lead members to treat each other’s contributions relatively equally. This may operate to reduce members’ willingness to put forth persistent and intense effort toward the task (Lawler 2005). Ultimately this may undermine group performance on disjunctive tasks. In contrast, in groups that are objectively diverse and adopt an individualistic norm, members are most likely to focus on genuine differences in their expertness levels, making it easier to identify important knowledge and improve group decisions, thereby providing the information elaboration and expertness identification needed for disjunctive task performance. In support of this thinking, Goncalo and Staw (2006) showed that individualistic norms are more beneficial than
Data and Sample The Himalayan mountain range—which stretches across Pakistan, India, Nepal, Tibet, and China—is home to the world’s most imposing peaks, including Mt. Everest, the tallest mountain on Earth. This unparalleled concentration of climbing challenges makes the Himalayas the most sought-after destination for accomplished mountaineers. We used the Himalayan Database (Salisbury and Hawley 2004 [2013]), which contains the detailed expedition records of Elizabeth Hawley, who since 1963 has served as the unrivalled chronicler of Himalayan expeditions (Jolly 2010, p. 1). Hawley has interviewed climbers from nearly all expeditions during the last half-century (Salisbury and Hawley 2004 [2013]). The Himalayan Database contains comprehensive information on 59,975 climbers who attempted Himalayan ascents in 8,184 expeditions between 1950 and 2013. We excluded several types of expeditions and climbers that did not fit with our study design, such as solo and two-climber ascents.^1 Our final sample consisted of 38,818 climbers in 5,214 expeditions. Dependent Variables Conjunctive Task Performance: Summiting success. Summiting the targeted peak is a critical performance outcome in mountain-climbing groups; it is the central accomplishment, the ultimate desired end state, and the very basis for the expedition’s formation. Consistent with Steiner’s (1972) typology of group tasks, summiting is a conjunctive task, which he defined (p. 17) as those in which “each member of a group is required to perform essentially the same function, and everyone’s success is determined by the effectiveness with which the least proficient member operates.” Steiner (1972, p. 28) cited mountain climbing specifically as the best example of this, given the extent to which the group’s overall performance can be constrained by the limitations of its weakest member. For example, reaching the summit of a target peak can require capitalizing on a temporary break in the weather, and delays incurred because of a lagging climber can reduce the likelihood of each climber making it to the summit (Krakauer 1997). In this context, developing a level of group solidarity that motivates climbers to stick together and in which members encourage one another to persist can significantly influence summiting success. As Connally (2004, p. 13), states, “little time is saved by letting groups spread out, because slow hikers get
even slower, and there’s always the danger of serious time loss if a party member goes missing.” We constructed a group-level, continuous measure of summiting success by calculating the proportion of climbers in each group who reached the summit of their target mountain ( x = 0.32, SD = 0.35).^1 Disjunctive Task Performance: Climber Safety****. Climber safety is another critical outcome for mountain-climbing groups and requires perpetual vigilance toward the environment and the constant re- calibrating of risk. Climber safety is maximized when mountaineering groups accurately detect and assess impending weather conditions and choose the route that is the safest. This requires surfacing and accurately weighting the expertness that is available within the group regarding its approach, particularly with respect to route selection, during each phase of the climb. As one mountain climbing textbook (Connally 2004, p. 15) describes it: “Routefinding may mean choosing from among several options for attaining your objective—a snowfield, a scree slope, or a pitch of technical rock; whether to climb a gully or a ridge; whether to follow a trail or go cross-country. The best route may depend on the season and weather, the condition of your party, the amount of daylight available, the equipment you’re carrying, and the consequences of misjudgment.” Connally (2004, p. 342) stresses the importance of expertness to this process: “Group decisions can easily be inferior to decision making by individuals, particularly in the face of uncertainty and ambiguity. Decisions taken often have more to do with relationships within the group (^1) Given our focus on summiting success, we excluded the 3,651 climbers who elected to only go as far as base camp, verified by climbing permits issued by the Nepalese government which require expeditions to identify in advance each climber who will be ascending above base camp. In addition, we excluded 872 climbers who were registered to attempt a summit but failed to reach base camp (183 expeditions included at least one such climber), leaving us with 55,452 climbers in 7,968 expeditions. Next, we excluded the 10,212 paid support personnel after calculating our control variable measuring climber support, which we describe below. This reduced our sample to 45,240 climbers in 7,962 expeditions (the reduction in number of expeditions at this stage is due to certain expeditions comprising entirely hired personnel such as all-Sherpa teams who were contracted to clear a trail in advance of commercial expeditions). Third, given our focus on group composition and consistent with typical definitions of a group as having three or more members (Kashy and Kenny 2000), we excluded 1,265 climbers who made ascents by themselves or with support personnel but no additional team members. This reduced our sample to 43,975 climbers in 6,697 expeditions. We also excluded 2,412 climbers who ascended in pairs, reducing our sample to 41,563 climbers in 5,491 expeditions. Next, because of the collectivism norm’s centrality to our analysis, we excluded all climbers from countries who are not included in Hofstede’s index (located at https://geert-hofstede.com/countries.html) as well as any expeditions that included at least one such climber. This reduced our number of expeditions to 5,244 and our number of climbers to 38,986. Finally, climber age, one of our control variables described below, was missing for 1,190 climbers, or approximately 3% of the remaining sample. These climbers were included in calculations of the control variable for the size of their group, but their age was coded as missing for the calculation of the control variable pertaining to their average expedition age. Age was missing for every climber in 30 expeditions comprising 168 climbers. Since we were unable to calculate an average age for these expeditions, we excluded them from our analysis. (However, we re-ran the analyses described below, assigning the mean group age (37.08 years) to those expeditions that were missing average age, and found no differences in the pattern of results.)
includes scores for 102 countries on a 100-point scale that ranges from 6 (Guatemala) to 91 (U.S.), with higher numbers representing lower collectivism. Hofstede based the Index on surveys of IBM employees in 64 countries, and then refined it through research on different work populations (Hofstede and Spangenberg 1987). Accordingly, we matched each expedition member’s country of citizenship to that country’s score on Hofstede’s Index.^2 This methodological approach has considerable precedent (Chua et al. 2015, Kalmijn and Tubergen 2010), including being used with Himalayan data specifically (Anicich et al. 2015). Our final sample included climbers from 80 countries, with the largest number of climbers (4,824) hailing from Japan while a single climber represented Bhutan, Morocco, Saudi Arabia, the United Arab Emirates, Tanzania, and Uruguay, respectively. We derived a continuous measure of collectivism for each expedition by averaging the climbers’ individualism scores. We then used the REVRS module in STATA to reverse these values so that higher numbers reflect greater collectivism. Next, we re-scaled these values from 0 to 1 by dividing each expedition’s collectivism score by 100 ( x = 0.30, SD = 0.21) to ensure comparability with our measure of relations-oriented diversity, described below. Finally, we mean- centered this variable in all analyses. Objective relations-based diversity. Objective national diversity is the extent to which each expedition included members from different nations. We used Blau’s (1977) index of heterogeneity, which measures the sum of squares of the proportion of expedition members from each nation: 1 −
(^2) Even though there is substantial precedent for assigning country-level Hofstede scores to individuals, some have criticized this practice, noting that within-country variance in behaviors related to dimensions like collectivism can be lost when assessed indirectly (Kirkman et al. 2006). To address this concern, we collected additional data from a sample of Himalayan climbers to examine whether country-level scores reflected individuals’ normative orientation with respect to collectivism. The sample consisted of 180 climbers from Himalayan expeditions between 2008 and 2013 (we asked participants to focus on their most recent expedition). Participants averaged 40 years old; 14% were female; and they had attempted an average of 1.43 Himalayan climbs. In addition, 44% used oxygen, while 51% participated in a commercial expedition. We asked these climbers to “Please rate each norm on a scale of 1 to 7, with 1 meaning the norm was highly uncharacteristic of your expedition and 7 meaning the norm was highly characteristic of your expedition.” These items were: “It was important to climbers that they respected the group’s collective decisions;” “Climbers stuck together, no matter what sacrifices were required;” “Climbers felt that it was their duty to take care of one another even if they had to sacrifice their own preferences or accomplishments;” and “Climbers on the expedition stayed together as much as they could.” We factor-analyzed responses using varimax rotation. One factor measuring collectivism emerged with an eigenvalue of 1.93. Each of the four items loaded over 0.60 on the factor without any cross-loadings and with a Cronbach’s alpha of .80 so we averaged responses to these four items to create a collectivism scale for each respondent. This measure, when entered into a regression analysis, significantly predicted climbers’ country-level collectivism score ( β = 3.71, p = 0.005), providing evidence that country-level scores based on Hofstede’s index were valid measures of collectivism at the individual level.
For example, an expedition with three climbers from Argentina and one from Norway would have a national diversity score of 0.38. The mean national diversity score across expeditions was 0.17 ( SD = 0.25), which we mean-centered in all analyses. Objective task-based diversity. We created a variable measuring dispersion in climbers’ objective expertness in the Himalayan region within each expedition, which is relevant given the uniquely challenging conditions. We measured expertness as the number of times a climber appears in the Himalayan Database prior to each focal climb ( x = 1.32, SD = 3.41). Sixty-one percent of ascents were attempted by a climber with no previous climbs in the region; 25% by a climber with one or two previous climbs in the region; and the remaining 14% were attempted by a climber with three or more previous climbs in the region. Ten percent of the climbers in our sample had more than three prior climbs, while less than 3% had 10 climbs or more.^3 We used these data to construct, for each expedition, the coefficient of variation in climber expertness ( x = 1.31, SD = 0.93). The coefficient of variation is the standard deviation of climber expertness divided by the mean of climber expertness in the expedition.^4 (^3) As a robustness check, we also constructed a Blau index for climber experience using the three categories of experience used for calculating our objective task diversity variable (0 prior climbs, 1-2 prior climbs, or three or more prior climbs). The mean of this index was 0.35 ( SD = 0.22). We estimated models with this covariate in place of the coefficient of variation reported above and obtained the same pattern of results. (^4) As is commonly noted in journalistic accounts (e.g., Jolly 2010), the Himalayan Database represents an unusually complete and accurate record—particularly with respect to climber names, generally including extra identifiers such as middle names and nicknames. Nevertheless, as with any database of this size, the potential for inaccuracies in record keeping exists. Accordingly, as a robustness check, we investigated any name-related discrepancies, due to the fact that we constructed our experience variables on the basis of climber names. We focused our attention on repeat climbers. The goal was to identify any instance of repeat climbers who may have, in actuality, been two different people, on the base of discrepant identifying information: Specifically, year of birth and current place of residence. We started by identifying the 458 repeat climbers who exhibited a standard deviation of greater than 0 in their year of birth. Many climbers were included in this category because at least one of their entries did not include a year of birth. For example, Aitor Iparragirre Sagarna appears twice in our final data, once with a 1972 year of birth and once with a missing year of birth—but, in both instances, showing the same residence. We did not consider climbers who fit this profile to be at risk. An additional type of climber had entries for year of birth in each appearance, but one or more discrepancies in the year itself. For example, Alix Christin Dorothee Von Melle appears five times in our final data. In each case, his residence is listed as Hoehenkirchen, Bavaria, Germany, and in all cases but one his year of birth is listed as 1971 (the exception is 1981). We did not consider climbers who fit this profile to be at risk. Our risk profile therefore consisted of repeat climbers who exhibited different years of birth and places of residence in different entries, though of course this could also reflect a combination of record-keeping errors and change of residence in between climbs. For example, Alberto Bianchi appears six times in our data; twice, his year of birth and place of residence are listed as 1943 and Como, Italy, while four times his year of birth and place of residence are listed as 1949 and Milano, Italy. Accordingly, we coded him as “at risk” for the purposes of this robustness check. In total, we identified 151 climbers who fit this “at risk” profile. We next coded their experience as missing and re- ran all the models depicted in Table 2. Doing so did not change our pattern of results.
