https://orcid.org/0009-0001-9580-6415
Figures
Abstract
This study examines peer effects among employees working fully remotely. We use panel data from a company that has operated with an entirely remote workforce since its inception and leverage as-if random assignment of new hires to teams as a quasi-natural experiment. We find no evidence that the average productivity of a worker’s teammates affects that worker’s own productivity. However, when team members are highly experienced, the productivity of employees on those teams increases by about 12.2%. In particular, employees with the shortest tenure see an increase in productivity of approximately 26.2%. Furthermore, this effect appears unrelated to the volume of communication within the team, suggesting that experienced teammates can have a positive influence even with minimal interaction, possibly through more efficient, targeted communication.
Citation: Ishikura H (2026) Experienced teammates increase productivity in remote work: Evidence from a full remote work company in Japan. PLoS One 21(3): e0342730. https://doi.org/10.1371/journal.pone.0342730
Editor: Nhat Tan Pham, International University - Vietnam National University Ho Chi Minh City, VIET NAM
Received: June 11, 2025; Accepted: January 27, 2026; Published: March 17, 2026
Copyright: © 2026 Hideaki Ishikura. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: The data used in this study are derived from internal human resource and operational records provided to the authors under a confidentiality agreement with Caster Inc. (Tokyo, Japan). These data contain sensitive personnel information and therefore cannot be shared publicly or deposited in a public repository. The authors do not have permission to distribute the underlying individual-level data. However, to ensure transparency and reproducibility, all analysis code and detailed variable construction scripts used in this study have been archived by the authors. The code will be made available to interested researchers upon reasonable request, subject to approval by Caster Inc. Requests for code should be directed to the corresponding author (Hideaki Ishikura; e-mail: ishikura.hideaki@keio.jp). The authors will coordinate internally with Caster Inc. as needed and will provide the code without undue delay, typically within a reasonable time frame. Access to the code does not require access to the confidential underlying data. Contact information: Caster Inc. Human Resources Department Otemachi First Square West Tower 1–2F LIFORK Otemachi R06, 1-5-1 Otemachi, Chiyoda-ku, Tokyo 100-0004, Japan Email: support@cast-er.com Website: https://caster.co.jp Citation: Caster Inc. (2024). Internal Human Resource Database. Tokyo, Japan.
Funding: The author(s) received no specific funding for this work.
Competing interests: One of the authors served as a director of Caster Inc. from October 2016 to November 2023 but has since resigned. The authors have a data use agreement with Caster Inc. to conduct academic research. This does not alter our adherence to PLOS ONE policies on sharing data and materials. The authors declare no other competing interests related to employment, consultancy, patents, or products in development.
Introduction
One of the major changes in work arrangements due to the COVID-19 pandemic is the rise of remote work; as of June 2023, 28% of workers in the United States and approximately 27% of workers in Japan had adopted remote work (Barrero et al. [1]; Ministry of Land [2]). Other surveys in 34 countries found that 25.6% of workers were engaged in hybrid work and 7.9% were working exclusively remotely as of April–May 2023, indicating that remote work has become an established mode of work in the post-pandemic era (Aksoy et al. [3]).
However, much of the existing research on remote work has focused on comparing productivity under remote versus on-site (or hybrid) arrangements. These macro-level studies document overall productivity differences across work arrangements but provide limited insight into within-team dynamics in remote settings. In particular, relatively few studies have directly examined whether peer effects—the influence of coworkers on an individual’s productivity—occur in fully remote teams. The limited evidence to date suggests that the productivity spillovers well documented in co-located offices may be muted or absent when employees work remotely. For example, Frakes and Wasserman [4] found that knowledge sharing among U.S. patent examiners boosted output only when peers worked in the same physical location, with this peer effect vanishing once examiners telecommuted. Similarly, Emanuel et al. [5] observed that engineers received substantially more feedback from colleagues in the same building than from those in different locations—a benefit that disappeared in a remote setting. Moreover, broad analyses indicate that remote work can sometimes reduce overall productivity by 12–20%, partly due to changes in worker interactions such as increased coordination delays and fewer spontaneous exchanges. Consistent with this view, Yang et al. [6] show that shifting to company-wide remote work reduced interpersonal connectivity and cross-team collaboration within a large technology firm. Taken together, these findings highlight an important gap: it remains unclear whether coworker-induced productivity gains can materialize when all team members are geographically dispersed, and if so, what mechanisms might enable such peer effects in fully remote settings.
By contrast, there is abundant evidence of peer effects and productivity spillovers in traditional in-person workplaces. Classic studies have shown that having high-performing or experienced colleagues can significantly raise an individual’s output. For example, Mas and Moretti [7] demonstrated that supermarket cashiers increased their checkout speed when more productive peers were present, highlighting the role of peer pressure. Likewise, Jackson and Bruegmann [8] and Bandiera et al. [9] found that employees working alongside particularly skilled or veteran coworkers achieved higher productivity, consistent with learning and knowledge spillover mechanisms. These findings underscore that coworker interactions in shared physical environments can enhance individual performance through both motivational pressure and on-the-job learning.
To frame how such peer effects might operate in remote teams, we draw on transactive memory system (TMS) theory. A TMS refers to a team’s shared understanding of “who knows what,” allowing members to locate and leverage each other’s expertise efficiently (Wegner [10]; Liang et al. [11]). In teams with a well-developed TMS, newcomers can more easily identify knowledgeable colleagues and obtain targeted guidance, which may enhance productivity even in the absence of physical proximity. While TMS and related mechanisms have been shown to matter in co-located teams, it remains an open question whether and how these mechanisms function in fully remote environments.
In light of this gap, this study examines whether working with more experienced teammates causally affects individual performance in a fully remote work setting. We use longitudinal personnel data from a Japanese company that has operated with an entirely remote workforce since its inception. New hires in this firm are effectively randomized into teams upon joining, which we exploit as a quasi-natural experiment to identify peer effects on productivity. Using this setting, we estimate the causal impact of team composition—particularly teammates’ average tenure and prior productivity—on individual output. We further examine heterogeneity by worker experience and explore whether patterns of intra-team communication are associated with productivity spillovers.
In preview, our analysis shows that having more experienced teammates significantly increases individual productivity in a fully remote setting, whereas having higher-prior-performance teammates does not yield comparable benefits. The productivity gains from experienced teammates are especially pronounced for workers with short tenure. By contrast, neither high coworker productivity nor a high volume of communication is associated with higher individual output, and for some experienced workers, lower communication intensity appears more conducive to productivity. These findings suggest that experience-based knowledge sharing, rather than peer pressure or frequent interaction per se, plays a central role in shaping productivity in remote teams.
The remainder of the paper is organized as follows. Section 2 reviews the related literature. Section 3 describes the data and empirical methodology. Section 4 presents the results, and Section 5 discusses the implications and concludes.
Literature review
Research on productivity under remote work arrangements is growing, but a clear consensus has yet to emerge. Bloom et al. [12] conducted a randomized controlled trial on call-center employees in China and reported that working from home increased productivity by around 13%. Other studies (e.g., Angelici and Profeta [13]; Choudhury et al. [14]) similarly find that remote work and flexible working arrangements can improve worker productivity. On the other hand, Gibbs et al. [15], Emanuel and Harrington [16], and Morikawa [17] reported that productivity under remote work fell by 12–20%. Gibbs et al. [15] suggest this decline is due to increased coordination tasks and online meetings reducing uninterrupted work time, whereas Emanuel and Harrington [16] attribute it to lower-quality workers entering the remote workforce.
It has also been noted that productivity in remote work varies with the nature of the task and the worker’s experience with remote work. Dutcher [18] found that the impact of working from home depends on task type: it boosted productivity by 11–20% for creative tasks but reduced it by 6–10% for monotonous tasks. Barrero et al. [1] and Okubo et al. [19] report that productivity does not decline—and may even improve—once workers have sufficient experience with remote work. Gibbs et al. [15] and Morikawa [17] make similar points.
Another line of research concerns transactive memory systems. Wegner [10] introduced TMS as a collective memory structure in which group members not only store knowledge individually but also maintain awareness of “who knows what.” Liang et al. [11] extended this concept to work teams and demonstrated experimentally that groups developing a TMS outperformed those that did not. Later work (Lewis [20]) confirmed that strong TMS improves coordination and team outcomes. In our context, we interpret teams with high average tenure as more likely to have developed a robust TMS, which can lower newcomers’ search costs for information and enhance their productivity. There is extensive research on peer effects among coworkers and between supervisors and subordinates in traditional settings. Falk and Ichino [21], Bandiera et al. [22], and Mas and Moretti [7] showed that peer pressure among colleagues positively affects productivity, especially for low-skilled workers in routine tasks. Many studies have also documented peer effects among highly skilled professionals. For example, Arcidiacono et al. [23], Song et al. [24], Chan et al. [25], De Grip and Sauermann [26,27], and Bandiera et al. [9] report that employees’ productivity improves when they work with highly capable and productive coworkers, attributing this to knowledge spillovers. Moreover, De Grip et al. [27] and Jackson and Bruegmann [8] showed that working with experienced coworkers positively impacts a worker’s productivity. Similarly, some studies have found spillovers between supervisors and subordinates:Lyle and Smith [28] and Lazear et al. [29] found that subordinates paired with high-performing bosses improve their own performance and have higher promotion rates. In summary, numerous studies indicate positive peer effects in co-located work environments, both among coworkers and between supervisors and subordinates.
Apart from mechanisms like peer pressure and “learning by watching,” many studies have highlighted the importance of communication for peer effects. This suggests that knowledge transfer often occurs through structured or informal communication among coworkers. Sandvik et al. [30] conducted an RCT in a sales company to test interventions that encourage communication-driven spillovers. They compared groups with different approaches: one encouraged employees to actively share know-how and socialize (e.g., have lunch together), another provided financial incentives for information sharing, and a third combined both. They found that the group encouraged to share knowledge and socialize had higher productivity, and this effect was long-lasting. This implies that reducing the “initial cost” of communication (i.e., the hesitation or uncertainty about asking questions) is crucial for facilitating knowledge sharing. Other studies echo this point (Catalini [31]; Cai and Szeidl [32]; Hasan and Koning [33]; Boudreau et al. [34]; Battiston et al. [35]; Englmaier et al. [36]).
On the other hand, research on peer effects in remote work environments is still in early stages, and many questions remain regarding the magnitude of effects and the mechanisms involved. Emanuel et al. [5] found that engineers with coworkers in the same building received 22% more online feedback than those with coworkers in different locations, but this difference vanished under remote work—highlighting the importance of physical proximity. Frakes and Wasserman [4] likewise reported that a one standard deviation increase in a coworker’s grant rate raised the grant rate of a junior patent examiner by 0.15 standard deviations, but this peer effect disappeared when examiners worked remotely. In academia, findings are mixed: Azoulay et al. [37] and Waldinger [38] found that intellectual closeness mattered more than physical proximity for knowledge spillovers, while Catalini [31] and Boudreau et al. [34] pointed to dramatically higher collaboration rates when researchers share the same location. Several studies suggest that communication and collaboration are key for spillovers in remote work (Atkin et al. [39]; Yang et al. [6]; Garro-Abarca et al. [40]; Jarrahi and Sawyer [41]). Kirkman et al. [42], Eisenberg et al. [43], and Leenders et al. [44] concluded that in virtual teams, both excessive and insufficient communication can hurt performance, and it remains unclear whether communication quality or quantity is more important.
As noted, while many studies examine the benefits of working with talented or experienced colleagues in co-located settings, relatively little research has focused on peer effects among coworkers in fully remote environments. Frakes and Wasserman [4] and Emanuel et al. [5] show that productivity spillovers observed in person tend to disappear under remote work arrangements. However, these studies were conducted during the sudden and large-scale transition to remote work during the COVID-19 pandemic and focused on specific occupations and organizations, limiting their external validity. Moreover, because they rely on comparisons between in-person and remote settings, they do not establish whether peer effects can arise among workers who are fully remote from the outset. In addition, although communication and collaboration are often viewed as central mechanisms for productivity spillovers in remote teams, existing research provides little guidance on which aspects of communication—such as quantity versus quality—are most relevant.
Taken together, the existing literature highlights well-established peer effects and productivity spillovers in co-located work environments, while leaving important questions unanswered in the context of fully remote work. In particular, it remains unclear whether peer effects can emerge among workers who are geographically dispersed from the outset, which types of coworkers matter most for individual performance, and through which mechanisms productivity spillovers may operate in remote teams. To address these unresolved issues, the next section develops a theoretical framework based on transactive memory system theory and the peer effects literature, and derives a set of testable hypotheses regarding how teammate experience, coworker productivity, and interaction patterns shape individual performance in fully remote settings.
Theoretical framework and hypotheses
Building on the literature above, we develop three hypotheses about how experienced teammates affect individual productivity in a fully remote work setting. Prior research on peer effects indicates that coworkers can significantly influence each other’s performance through mechanisms of social pressure, learning, and knowledge sharing. In remote teams, where face-to-face contact is absent, these mechanisms may operate differently, but we expect that experienced colleagues still provide important productivity benefits. We outline our hypotheses below, grounded in relevant theory and empirical findings.
Hypothesis 1 (H1): Having more experienced teammates increases an individual’s productivity. This hypothesis follows from evidence of positive peer effects in traditional workplaces and the knowledge-transfer advantages of experienced coworkers. Economists have documented that introducing high-performing, experienced workers into a team can raise the productivity of others through peer influence. For example, Mas and Moretti [7] found strong productivity spillovers when highly productive personnel were added to a shift of grocery workers. Similarly, Bandiera et al. [9] showed that fruit pickers worked about 10% faster when paired with a friend who was more able (productive) than themselves. These studies suggest that experienced or high-ability teammates set performance norms and directly or indirectly encourage greater effort from their peers. In addition to social pressure, experienced teammates are likely to possess valuable task knowledge that can be shared. According to transactive memory theory, team members develop a shared “who knows what” knowledge system (Wegner [10]) and draw on each other’s expertise. Empirical work has linked such collective knowledge systems to better group performance. Thus, both social influence and knowledge-sharing perspectives predict that working with more experienced colleagues will improve an individual’s output. We therefore hypothesize that a worker’s productivity rises with the average experience level of their teammates (H1).
Hypothesis 2 (H2): The productivity gains from experienced teammates are larger for less experienced (junior) workers than for more experienced workers. We posit a moderating effect of the focal worker’s own experience. Less experienced employees have greater scope for improvement and more to learn, so they should benefit disproportionately from teammates who can offer guidance, expertise, and norms of high performance. In contrast, veteran employees might already possess similar knowledge or may be less impressionable to peer influence. This idea is consistent with social learning arguments and findings in the peer-effect literature. For instance, Bandiera et al. [9] observed that in a workplace setting, employees who were less able than their friends significantly increased their productivity when a more able friend was present. In essence, junior workers may “borrow” expertise and best practices from seasoned colleagues, leading to larger relative gains in productivity. Therefore, we expect the positive effect of experienced teammates to be more pronounced for relatively inexperienced workers (H2).
Hypothesis 3 (H3): The positive impact of experienced teammates on productivity is stronger when the teammates have a history of frequent interaction (greater familiarity). Here we consider the role of team familiarity and repeated collaboration. Prior research suggests that peer effects are not uniform – they intensify when coworkers regularly interact. Mas and Moretti [7] found that workers responded more strongly to high-productivity peers with whom they frequently interacted, implying that familiarity or repeated contact amplifies peer influence. Theoretically, this can be explained by the development of transactive memory and coordination routines over time. Through repeated interaction, team members learn “who knows what” and how to work together efficiently (Wegner [10]), leading to a shared knowledge repository and better coordination. Liang et al. [11] provide empirical support for this: groups that trained together (developing shared memory of each other’s expertise) outperformed those that did not, underscoring the performance benefits of established team knowledge systems. In a remote work context, building familiarity might require deliberate communication, but the underlying premise remains that teammates who have collaborated more often can transfer know-how and tacit knowledge more effectively. Accordingly, we hypothesize that an experienced teammate’s positive effect on an individual’s productivity will be amplified when that individual has had more frequent or longer-term interactions with the teammate (H3).
Data and methods
Overview of Caster Inc. and analytical sample
The company has operated with a fully remote workforce since its inception; as of March 2025, it employed approximately 900 people across all 47 prefectures of Japan and in 23 other countries. This study analyzes employees in the CASTER BIZ department. We use de-identified individual-level administrative and personnel records covering April 2022 to March 2023.
New hires at Caster undergo a six-month on-the-job training (OJT) period, during which they often do not serve clients independently. To ensure that our sample consists of workers fully engaged in client-facing tasks, we exclude all employee-month observations where tenure was less than six months. The final analytical sample includes 977 employee-month observations. This dataset reflects the full population of employees in the CASTER BIZ department who met the inclusion criteria during the study period.
Data collection procedures
Access and permissions. Data were obtained from Caster Inc. under a formal data use agreement for academic research. All data were anonymized prior to analysis: direct identifiers were removed or masked, and each employee was assigned a unique, non-identifiable ID. Access to the raw data was limited to the research team.
Data sources. The dataset integrates three internal systems: (1) the human resources management system; (2) operational performance databases; and (3) Slack communication logs. The HR system provided demographic and employment data; operational databases provided monthly client assignment counts and work-hour records; and Slack logs provided internal communication volume.
Data extraction and processing. Monthly records from each source were extracted for April 2022 through March 2023, cleaned, checked for consistency, and merged into an employee-month panel using unique employee identifiers. Observations with tenure under six months were excluded to remove the training period.
Variable construction. Productivity is measured by the number of unique clients handled per employee per month, recorded automatically in the client management system. Communication volume is measured by the number of internal Slack messages per employee per month (excluding client-facing messages). Team-level variables (e.g., average tenure, productivity, and communication volume) are computed as the average for other team members, excluding the focal individual. “Top” and “bottom” categories for experience or communication are defined using the top and bottom 25% of the relevant distributions. All definitions and cutoffs were specified prior to empirical analysis.
Institutional setting
Caster Inc. provides online “secretary” services. Customers who contract with Caster can request various back-office tasks—such as secretarial work, human resources, accounting, and general affairs—from online secretaries (Caster employees) for up to 30 hours per month. Clients primarily submit requests via online chat tools (e.g., Slack or Microsoft Teams), and the online secretary carries out the requested work.
Employees hired as online secretaries undergo a six-month on-the-job training period, after which they are assigned to clients. The company follows a “same work, same pay” principle: an online secretary’s monthly salary is determined by the number of clients they serve, regardless of their previous salary. There is no minimum quota of clients; each employee can decide how many clients to take on.
New online secretaries are allocated by a mechanical round-robin rule: the first newcomer goes to Team A, the next to Team B, then Team C, and so on, cycling back to Team A. Neither managers nor workers can deviate from this rotation—there is no screening, swapping, or requesting of teams—so neither managers nor new hires have discretion over the destination team. Because the order of arrivals is orthogonal to team characteristics, the resulting placement is as-if randomized across teams within each join month. We exploit this institutional feature as a quasi-natural experiment in our analysis.
Each team is designed to consist of 4–12 online secretaries and exactly one designated team leader. Team leaders do not have their own client quotas and there are no additional performance incentives based on team outcomes. When constructing our analytical sample, we verified that all teams in the study period followed this “single-leader” structure. Any teams that would not have met this criterion (e.g., teams without a leader or with multiple leaders) would have been excluded, although in practice no such cases occurred in our dataset.
Outcome and key variables
The primary outcome variable is individual productivity, proxied by the number of clients handled in a given month. This is a reasonable proxy because the firm operates under standardized working hours (9:00–17:00 on weekdays), and each client contract entitles the client to 30 hours of monthly support. The more clients a worker handles, the greater the implied output, assuming roughly comparable working hours across employees.
Key explanatory variables include the lagged average productivity and average tenure of teammates (excluding the individual), and indicators for whether a worker’s team falls into the top or bottom quartile of experience or communication intensity. Communication volume is derived from internal Slack messages among employees.
Empirical strategy
To estimate the impact of team characteristics on individual productivity, we estimate fixed-effects panel models that control for time-invariant individual heterogeneity and common time shocks. The key independent variable is the one-month lag of teammates’ average productivity (or experience), excluding the focal individual. Control variables include age, gender, tenure, team size, working time, and lagged average team Slack activity (excluding the individual). All models include individual, team, and calendar-month fixed effects. Standard errors are clustered at the individual and team levels.
where subscripts ,
, and
denote individual, team, and month, respectively.
is the outcome variable, defined as the natural logarithm of the number of clients that individual 𝑖 in
team handles at time
.
is the average number of clients handled by team
at time
, excluding individual 𝑖.
is a vector of control variables, including individual
’s and team
’s characteristics (age, female, team size, tenure, working time and post team mean at time
, excluding
).
is an individual fixed effect capturing time-invariant characteristics of worker 𝑖,
is a team fixed effect,
is a month fixed effect, and
is an error term.
When estimating peer effects, two issues must be addressed. First, we must account for potential endogeneity related to individual and team characteristics. For example, unobserved abilities of individuals or their coworkers could influence productivity, biasing the estimates. We mitigate this by including individual and team fixed effects in the model. Second, there is the reflection problem (Manski [45]), which complicates causal interpretation because coworkers influence each other simultaneously.
To address the reflection problem, Yeung and Nguyen-Hoang [46]suggest two approaches: using instrumental variables in a two-stage least squares framework, or using lagged peer variables. Following prior studies (Yeung & Nguyen-Hoang [46]; An et al. [47]; Cassar and Ko [48]), we use the one-month lag of peer average productivity (team average clients excluding 𝑖) to mitigate simultaneity.
Robustness and randomization checks
We leverage the firm’s mechanical team assignment system (round-robin by hire order) to claim as-if random assignment of new hires to teams. To support this claim, we conduct ANOVA tests and permutation-based randomization inference. We further test robustness by including team-by-entry-month fixed effects and alternative peer exposure definitions (e.g., indicator for at least one top-quartile peer). Results are consistent across specifications, confirming the validity of our identification strategy.
Descriptive statistics
Table 1 presents basic summary statistics for the sample. Detailed descriptions of variables are provided in Table 2.
Ethics statement
This study was an observational analysis of retrospective, fully anonymized business records from company employees. In accordance with institutional guidelines, formal ethics committee (Institutional Review Board) review was not required since the research did not involve any identifiable personal data or direct interaction with human participants. The dataset did not include any minors, and informed consent was not required because all records were pre-existing and de-identified prior to analysis. The study was conducted in compliance with all relevant ethical standards for research using anonymized data.
Results
Main estimation
Using the model above, we estimate the effect of team productivity on individual productivity. The results are shown in Table 3.
We find no evidence that being on a team with high average productivity raises an individual’s own productivity. Even for teams with top-quartile average productivity, we find no significant effect on individual output. However, when team members are highly experienced (teams with top-quartile average tenure), the productivity of workers on those teams increases by about 12.2%. In a fully remote work setting, we do not observe a peer effect from working with highly productive (high-output) coworkers, but we do find evidence of a peer effect from working with experienced coworkers.
Subsample estimation
To examine how peer effects vary with tenure, we split the sample by employee tenure quartiles and re-estimate the model for each subgroup (see Table 4).
The subsample results indicate that productivity increases by approximately 26.2% for workers in the shortest tenure group (first quartile), and by about 8.6% for those in the second quartile (tenure between the 25th and 50th percentile). These findings are consistent with Frakes and Wasserman [4], who observed peer effects for employees with less than two years at the company, with the effect diminishing as tenure increased.
By contrast, when we split the sample by individual productivity levels, we do not find statistically significant peer effects in any productivity quartile (see Table 5).
Additionally, we examined whether having exceptionally productive coworkers (top 25% individually) influences outcomes. We conducted subsample analyses by worker productivity and by tenure for those workers whose teams included highly productive peers, but found no significant effects in those cases either (see Tables 6 and 7).
These results suggest that in a remote work environment, working with experienced coworkers (as opposed to simply high-output coworkers) is what boosts individual productivity—particularly for employees who are new to the company.
The effect of communication
We next explore whether the volume of communication within a team influences individual productivity. The employees covered in this study do not communicate via email, hold only one company-wide meeting per month, and conduct nearly all communication through Slack. This study measures only the volume of internal Slack messages, excluding client-facing messages. We define this as communication volume and aim to clarify the effect of communication volume within teams on individual employee productivity. We compare teams with high average tenure (where we found peer effects) to other teams in terms of communication volume. Fig 1 shows that teams with top-quartile average tenure actually have a lower volume of communication than other teams. A Wilcoxon rank-sum test confirms a statistically significant difference (p = 0.001) in communication volume: high-tenure teams communicate less than other teams. This suggests that a high amount of intra-team communication might not be necessary for peer productivity effects to occur.
This figure shows the density function of communication volume on Slack between employees belonging to Top Tenure and those who do not. The former are shown in blue, the latter in red.
We also compared communication levels between teams with high versus low average productivity, and found no significant difference (see Fig 2). A Wilcoxon rank-sum test indicates no statistical difference in the volume of communication between the two (P-value = 0.486).
This figure shows the density function of communication volume on Slack between employees belonging to Top Pro and those who do not. The former are shown in blue, the latter in red.
Interestingly, when we segment by worker tenure and productivity, we find that long-tenured workers and already high-productivity workers tend to perform better when they are on teams with low communication levels. Table 8 and 9 present these results. In other words, experienced and highly productive employees seem to benefit, at least to some extent, from being in a quieter team environment.
Random assignment check
We exploit the random assignment of new employees to teams to help identify peer effects. A potential concern is that team assignments might be endogenous. To verify the randomness of team assignment, we conducted an ANOVA comparing characteristics of employees across teams.
The ANOVA results (see Table 10) indicate that only one characteristic—age—differs significantly across teams (at the 5% level).
We include age as a control in the regressions, which should account for this slight discrepancy. To address cohort-specific baseline sorting, we add team × entry-month fixed effects; results are virtually unchanged relative to the baseline, supporting the view that initial assignment is not driving our estimates. Identification then comes from within team × cohort variation over time. As shown in Table 11, the results are virtually unchanged relative to the baseline, supporting the view that initial assignment is not driving our estimates and is consistent with quasi-random assignment at entry.
Additionally, using randomization inference, we randomly reassigned team labels within entry month 300 times while preserving team sizes and re-estimated the same specification each time. The permutation p-value for the coefficient on the lagged peer average (team mean excluding the focal individual) was 0.193 (two-sided). The observed estimate is not unusually large relative to the randomization distribution, again consistent with quasi-random assignment at entry.
Robustness check
We conducted several additional analyses to check the robustness of our findings. First, we confirmed that there are no significant differences in observable worker attributes between teams with high vs. low average tenure, or between teams with high vs. low average productivity (see Table 12). This suggests that selection on observables is unlikely to drive our results.
We also tested alternative definitions of peer exposure. Instead of using team averages, we created indicators for whether a team contains at least one exceptionally high-performing coworker (top 25% in individual productivity) or at least one very experienced coworker (top 25% in tenure). We then examined the effects of these indicators on individual productivity (see Table 13). The results are in line with our main findings, confirming that our conclusions are robust. However, the estimated effect sizes using these indicators are somewhat larger (about a 30.1% increase for the shortest-tenure group), suggesting a possible upward bias when not fully accounting for individual and team fixed effects.
Discussion and conclusion
Discussion and conclusion
Our findings provide clear evidence that experienced teammates can significantly enhance individual productivity in a fully remote work setting, whereas simply working alongside highly productive coworkers does not. In particular, employees assigned to teams with high average tenure experienced substantial productivity gains, while no comparable spillover was observed from teammates’ average productivity or from higher communication volume. These results highlight the importance of experience-based mechanisms, rather than peer pressure or frequent interaction per se, in shaping productivity in remote teams.
Hypothesis 1 (H1), which predicted that working with more experienced teammates would increase individual productivity, is strongly supported by the data. Employees in teams with longer average tenure exhibited significantly higher output, especially among those with short tenure themselves. This pattern is consistent with transactive memory system (TMS) theory, which emphasizes the role of shared knowledge about “who knows what” within a team (Wegner [10]; Liang et al. [11]). In a fully remote environment, where informal observation and spontaneous interactions are limited, experienced teammates may help newcomers navigate tasks more efficiently by lowering search costs for information and providing targeted guidance. Our findings suggest that such TMS-based mechanisms can operate even in the absence of physical proximity.
By contrast, Hypothesis 2 (H2), which posited productivity spillovers from working alongside highly productive peers through peer pressure, is not supported. Unlike in co-located settings studied by Mas and Moretti [7] or Bandiera et al. [9], we find no evidence that teammates’ average productivity raises individual output in remote teams. One plausible explanation is that remote workers cannot easily monitor each other’s effort or performance, weakening social pressure. In addition, under the compensation system studied here, individual pay depends solely on personal output rather than team performance. As emphasized by Hamilton et al. [49], peer effects driven by peer pressure are more likely to arise when team-based incentives are present. In our setting, workers have little incentive to adjust effort in response to high-performing peers, which may explain the absence of peer-pressure effects.
Hypothesis 3 (H3), which emphasized knowledge spillovers mediated by interaction and communication, is also not supported in a straightforward manner. We find no evidence that higher volumes of communication enhance individual productivity, nor that working with highly productive teammates generates spillovers through frequent interaction. Instead, communication quantity appears to be a poor proxy for effective knowledge transfer in remote teams. This finding is consistent with prior work suggesting that excessive communication can crowd out focused work time (Bloom et al. [12]; Gibbs et al. [15]). At the same time, our results do not imply that communication is unimportant; rather, they suggest that the quality and targeting of interactions—potentially facilitated by experienced teammates—may matter more than sheer volume.
Importantly, our results distinguish clearly between the role of experience and the role of communication volume. While prior literature often treats communication as the primary channel of spillovers, our findings suggest that productivity gains stem from accumulated team experience rather than the sheer quantity of interaction. In other words, it is not frequent communication itself, but the presence of experienced teammates who can provide targeted and efficient guidance, that drives the observed productivity effects.
Taken together, these findings indicate that peer effects in fully remote teams differ fundamentally from those observed in co-located environments. While peer pressure and “learning by watching” play an important role in traditional workplaces, experience-based knowledge structuring appears to be the dominant mechanism in remote settings. Experienced teammates may provide just enough guidance to reduce the initial cost of asking questions for newcomers, without requiring frequent interaction. This interpretation aligns with evidence that experience is critical for sustaining productivity in remote work arrangements (Barrero et al. [1]; Morikawa [17]; Okubo et al. [19]).
Several limitations warrant discussion. First, although our fixed-effects strategy mitigates concerns about time-invariant heterogeneity, we cannot fully eliminate the possibility that unobserved time-varying characteristics influence both peer composition and productivity. Second, our analysis is based on a single fully remote organization, which may limit external validity. Third, we do not observe the content or quality of communication, preventing direct identification of knowledge-transfer channels.
In conclusion, our results show that who one works with continues to matter in remote work—but in different ways than in traditional offices. Assigning less experienced workers to teams with seasoned colleagues can yield substantial productivity gains, even in fully remote organizations. By contrast, simply grouping workers with high performers or increasing communication intensity does not generate comparable benefits. These findings suggest that effective team design in remote work should prioritize experience-based knowledge structures over peer pressure or frequent interaction.
Practical implications
These findings provide valuable insights for managers and organizations. Working in a team with experienced coworkers can improve the productivity of remote workers, especially those who are new to the company. In contrast, for employees who are themselves experienced or highly productive, an environment with fewer distractions and less frequent communication may be more conducive to productivity.
First, managers might consider team reassignments or mentoring setups based on employee experience and performance. For example, pairing junior, less-experienced employees with veteran coworkers could help accelerate learning and boost productivity by lowering the “initial cost” of seeking help. Meanwhile, giving more autonomy and quiet space to senior or high-performing employees—rather than placing them in communication-heavy teams—could help maintain or improve their output.
More broadly, as remote work continues to evolve, it is important for organizations to identify and implement evidence-based practices to support productivity. This will require leveraging data on employee performance and collaboration from multiple companies to pinpoint what works and why. Currently, many HR decisions are guided by intuition and experience; going forward, firms should invest in collecting and analyzing employee data, developing HR professionals who can use data for strategic decisions, and building data infrastructure to facilitate evidence-based management in a remote work era.
References
- 1. Barrero JM, Bloom N, Davis SJ. The evolution of work from home. J Econ Perspect. 2023;37(4):23–49.
- 2.
Ministry of Land, Infrastructure, Transport and Tourism. FY2021 Telework Population Survey. 2022. Available from: https://www.mlit.go.jp/toshi/daisei/content/001471979.pdf
- 3. Aksoy CG, Barrero JM, Bloom N, Davis SJ, Dolls M, Zarate P. Working from home around the globe: 2023 report. EconPol Policy Brief. 2023.
- 4. Frakes MD, Wasserman MF. Knowledge spillovers, peer effects, and telecommuting: Evidence from the U.S. Patent Office. Journal of Public Economics. 2021;198:104425.
- 5.
Emanuel N, Harrington E, Pallais A. The power of proximity to coworkers: training for tomorrow or productivity today? Cambridge (MA): National Bureau of Economic Research. 2023.
- 6. Yang L, Holtz D, Jaffe S, Suri S, Sinha S, Weston J, et al. The effects of remote work on collaboration among information workers. Nat Hum Behav. 2022;6(1):43–54. pmid:34504299
- 7. Mas A, Moretti E. Peers at work. Am Econ Rev. 2009;99(1):112–45.
- 8. Jackson CK, Bruegmann E. Teaching students and teaching each other: the importance of peer learning for teachers. Am Econ J Appl Econ. 2009;1(4):85–108.
- 9. Bandiera O, Barankay I, Rasul I. Social incentives in the workplace. Rev Econ Stud. 2010;77(2):417–58.
- 10.
Wegner DM. Transactive memory: A contemporary analysis of the group mind. In: Mullen B, Goethals GR, editors. Theories of group behavior. New York: Springer-Verlag. 1986:185–208.
- 11. Liang DW, Moreland RL, Argote L. Group versus individual training and group performance: The mediating role of transactive memory. J Pers Soc Psychol. 1995;69(5):865–78.
- 12. Bloom N, Liang J, Roberts J, Ying ZJ. Does Working from Home Work? Evidence from a Chinese Experiment *. The Quarterly Journal of Economics. 2014;130(1):165–218.
- 13. Angelici M, Profeta P. Smart working: work flexibility without constraints. Manage Sci. 2024;70(3):1680–705.
- 14. Choudhury P, Foroughi C, Larson B. Work-from-anywhere: The productivity effects of geographic flexibility. Strateg Manage J. 2021;42(4):655–83.
- 15. Gibbs M, Mengel F, Siemroth C. Work from home & productivity: Evidence from personnel & analytics data on IT professionals. J Polit Econ Microecon. 2023;1(1):7–41.
- 16. Emanuel N, Harrington E. Working remotely? Selection, treatment, and the market for remote work. Am Econ J Appl Econ. 2024;16(4):528–559.
- 17. Morikawa M. Productivity dynamics of remote work during the COVID-19 pandemic. Ind Relat. 2023;62(3):317–31.
- 18. Dutcher EG. The effects of telecommuting on productivity: An experimental examination. The role of dull and creative tasks. J Econ Behav Organ. 2012;84(1):355–63.
- 19. Okubo T, Inoue A, Sekijima K. Teleworker Performance in the COVID-19 Era in Japan. Asian Economic Papers. 2021;20(2):175–92.
- 20. Lewis K. Knowledge and performance in knowledge-worker teams: A longitudinal study of transactive memory systems. Manage Sci. 2004;50(11):1519–33.
- 21. Falk A, Ichino A. Clean Evidence on Peer Effects. Journal of Labor Economics. 2006;24(1):39–57.
- 22. Bandiera O, Barankay I, Rasul I. Social preferences and the response to incentives: Evidence from personnel data. Q J Econ. 2005;120(3):917–62.
- 23. Arcidiacono P, Kinsler J, Price J. Productivity Spillovers in Team Production: Evidence from Professional Basketball. Journal of Labor Economics. 2017;35(1):191–225.
- 24. Song H, Tucker AL, Murrell KL, Vinson DR. Closing the productivity gap: Improving worker productivity through public relative performance feedback and validation of best practices. Manag Sci. 2018;64(6):2628–49.
- 25. Chan TY, Li J, Pierce L. Compensation and peer effects in competing sales teams. Manage Sci. 2014;60(8):1965–84.
- 26. De Grip A, Sauermann J. The effects of training on own and co-worker productivity: evidence from a field experiment. Econ J. 2012;122(560):376–99.
- 27. De Grip A, Sauermann J, Sieben I. The role of peers in estimating tenure-performance profiles: evidence from personnel data. J Econ Behav Organ. 2016;126:39–54.
- 28. Lyle DS, Smith JZ. The Effect of High-Performing Mentors on Junior Officer Promotion in the US Army. Journal of Labor Economics. 2014;32(2):229–58.
- 29. Lazear EP, Shaw KL, Stanton CT. The Value of Bosses. Journal of Labor Economics. 2015;33(4):823–61.
- 30. Sandvik JJ, Saouma RE, Seegert NT, Stanton CT. Workplace knowledge flows. Q J Econ. 2020;135(3):1635–80.
- 31. Catalini C. Microgeography and the direction of inventive activity. Manage Sci. 2018;64(9):4348–64.
- 32. Cai J, Szeidl A. Interfirm relationships and business performance. Q J Econ. 2018;133(3):1229–82.
- 33.
Hasan S, Koning R. Conversational peers and idea generation: Evidence from a field experiment. Harvard Business School Strategy Unit. 2017.
- 34. Boudreau KJ, Brady T, Ganguli I, Gaule P, Guinan E, Hollenberg A, et al. A Field Experiment on Search Costs and the Formation of Scientific Collaborations. Rev Econ Stat. 2017;99(4):565–76. pmid:29375163
- 35. Battiston D, Blanes i Vidal J, Kirchmaier T. Face-to-face communication in organizations. Rev Econ Stud. 2021;88(2):574–609.
- 36.
Englmaier F, Grimm S, Grothe D, Schindler D, Schudy S. The effect of incentives in non-routine analytical team tasks. 2018.
- 37. Azoulay P, Graff Zivin JS, Wang J. Superstar extinction. Q J Econ. 2010;125(2):549–89.
- 38. Waldinger F. Peer Effects in Science: Evidence from the Dismissal of Scientists in Nazi Germany. The Review of Economic Studies. 2011;79(2):838–61.
- 39.
Atkin D, Chen MK, Popov A. The returns to face-to-face interactions: knowledge spillovers in Silicon Valley. Cambridge (MA): National Bureau of Economic Research. 2022.
- 40. Garro-Abarca V, Palos-Sanchez P, Aguayo-Camacho M. Virtual Teams in Times of Pandemic: Factors That Influence Performance. Front Psychol. 2021;12:624637. pmid:33679543
- 41. Jarrahi MH, Sawyer S. Social technologies, informal knowledge practices, and the enterprise. J Organ Comput Electron Commer. 2013;23(1–2):110–37.
- 42. Kirkman BL, Rosen B, Tesluk PE, Gibson CB. The impact of team empowerment on virtual team performance: The moderating role of face-to-face interaction. Acad Manage J. 2004;47(2):175–92.
- 43. Eisenberg J, Post C, DiTomaso N. Team Dispersion and Performance: The Role of Team Communication and Transformational Leadership. Small Group Research. 2019;50(3):348–80.
- 44. Leenders RTA, Van Engelen JM, Kratzer J. Virtuality, communication, and new product team creativity: a social network perspective. J Eng Technol Manag. 2003;20(1–2):69–92.
- 45. Manski CF. Identification of endogenous social effects: The reflection problem. Rev Econ Stud. 1993;60(3):531–42.
- 46. Yeung R, Nguyen-Hoang P. Endogenous peer effects: Fact or fiction? J Educ Res. 2016;109(1):37–49.
- 47. An W, Beauvile R, Rosche B. Causal network analysis. Annu Rev Sociol. 2022;48(1):23–41.
- 48. Cassar G, Ko T. Peer effects in subjective performance evaluation. Contemp Account Res. 2023;40(3):1704–32.
- 49. Hamilton BH, Nickerson JA, Owan H. Team incentives and worker heterogeneity: An empirical analysis of the impact of teams on productivity and participation. J Polit Econ. 2003;111(3):465–97.