Balancing Turnover and Promotion Outcomes: Evidence on the Optimal Hybrid-Work Frequency

TL;DR

The paper investigates the optimal frequency of hybrid/remote work by analyzing large-scale attendance data to assess turnover risk and promotion likelihood. Using a generalized propensity score approach and dose-response curves, it identifies a U-shaped turnover response and an inverted U-shaped promotion response, with an optimal hybrid frequency around two remote days per week. The study further reveals moderation by gender, job type, and leadership, and links these outcomes to time-allocation mediators such as working hours, schedule flexibility, and after-hours work. The findings offer actionable guidance for employers and policymakers to implement balanced hybrid policies that minimize turnover while maximizing promotion potential, while highlighting limits due to its short-term, single-company scope.

Abstract

Hybrid work policy, especially return-to-office requirements, remains a globally salient topic as workers, companies, and governments continue to debate and disagree. Despite extensive discussions on the benefits and drawbacks of remote and hybrid arrangements, the optimal number of remote days that jointly considers multiple organizational outcomes has not been empirically established. Focusing on two critical career outcomes -- turnover risk and promotion -- we examine how remote work frequency shapes employee trajectories using large-scale observational activity data from a company with over one million employees. We find that increased remote-work frequency is associated with an initial decrease and then an increase in turnover, while promotion likelihood initially rises and then declines. Accordingly, we identify approximately two remote days per week as an optimal balance -- maximizing promotion, a positive outcome for employees, while minimizing turnover, which is undesirable for organizations and may indicate negative employee experiences. These patterns vary across subgroups defined by gender, role type, and leadership status. Several notable results emerge. First, male employees derive greater promotion benefits from remote work than female employees. Second, support workers (non-core business roles) do not experience promotion gains, and the reduction in turnover at their optimal remote-work frequency is marginal compared with employees in core business roles. Third, organizational leaders face greater challenges in remote settings than individual contributors: their turnover risk increases substantially at higher remote frequencies, and their likelihood of promotion decreases as remote frequency rises. We further show that time-allocation patterns partly explain how remote-work frequency influences these career outcomes.

Figures (5)

• Figure 1: Dose-response curves with 95% confidence intervals for the weighted regressions of turnover risk on remote frequency. As remote frequency increases, turnover risk initially decreases, reaching its lowest point at 1.6 remote days per week, and then begins to rise. Regressions on subgroups defined by gender, position type, and leadership status reveal heterogeneous effects. (a): The optimal remote frequency for minimizing turnover risk is 1.4 days per week for men and 1.8 days per week for women. (b): For main business workers, the optimal frequency is 1.8 days per week, while for support workers it is 1.1 days per week. See the definitions of position types in the main body. (c): For individual contributors, the optimal frequency is 1.7 days per week, whereas for leaders it is 1.3 days per week.
• Figure 2: Dose-response curves with 95% confidence intervals for the weighted regressions of rank change on remote frequency. As remote frequency increases, rank change initially rises, peaking at 2.1 remote days per week, and then begins to decline. Regressions on subgroups defined by gender, position type, and leadership status reveal heterogeneous effects. (a): The optimal remote frequency for maximizing promotion is 2.3 days per week for men and 2.0 days per week for women. (b): For main business workers, the optimal frequency is 2.1 days per week, while the effect for support workers is limited. (c): For individual contributors, the optimal frequency is 2.3 days per week, while leaders achieve the highest promotion likelihood when working fully on-site.
• Figure 3: Instantaneous indirect effects (with 95% confidence interval) of remote frequency through working hours. (I) Working hours (scaled and vertically offset for clarity) increase with remote frequency. (II) Working hours do not significantly mediate the effect of remote frequency on turnover. (III) Increasing working hours strengthen the positive effect of remote work on promotion when remote frequency is 2.5 days or fewer per week. At higher remote frequencies, the indirect effect becomes non-significant, even though working hours continue to increase.
• Figure 4: Instantaneous indirect effects (with 95% confidence interval) of remote frequency through flexibility.(a): (I) Weekly schedule flexibility increases with remote frequency up to 1.5 days per week, then decreases, reaching its lowest point under fully remote conditions. (II) Flexibility strengthens the effect of remote frequency on turnover: when remote frequency is below 1 day per week, increased flexibility helps reduce turnover risk; when remote frequency exceeds 1.5 days per week, decreasing flexibility contributes to higher turnover risk. (III) At low remote frequencies ($<$ 1.5 days/week), increased flexibility buffers the positive effect of remote work on promotion. At moderate frequencies (2–3.5 days/week), declining flexibility mediates a positive effect. At high frequencies (4–5 days/week), flexibility drops below onsite levels, and the mediation effect becomes non-significant. (b) (I) Daily schedule flexibility (measured by the entropy of working hours) increases with remote frequency up to 2 days per week, then decreases. (II) Flexibility strengthens the effect of remote frequency on turnover: increased flexibility reduces turnover risk at remote frequencies up to 1.5 days per week, while at frequencies above 2.5 days per week, decreasing flexibility contributes to reduced turnover. (III) Flexibility also strengthens the effect of remote frequency on promotion at both low and high ends. When remote frequency is below 0.5 days per week, increased flexibility positively mediates the effect of remote work; at high remote frequencies ($\geq$ 4 days/week), flexibility is even lower than fully onsite conditions and negatively mediates the effect of remote work.
• Figure 5: Instantaneous indirect effects (with 95% confidence interval) of remote frequency through after-hour work.(a): (I) Weekend work (measured by the proportion of working days falling on weekends) decreases with remote frequency until 3.5 days/week, then slightly increases. (II) Weekend work buffers the effect of remote work on turnover: for remote frequencies up to 3 days/week, reduced weekend work mediates a positive effect; at higher frequencies (4–5 days/week), a slight increase in weekend work helps reduce turnover risk. (III) Weekend work strengthens the effect on promotion: reduced weekend work ($\leq$ 3 days/week remote) contributes to a higher chance of promotion, while increased weekend work at higher remote frequencies shows a negative mediation effect. (b): (I) Late work (measured by the proportion of working days with evening activity) decreases with remote frequency until 2 days/week, then increases. (II) Late work does not significantly mediate the effect of remote work on turnover. (III) Late work strengthens the effect on promotion: decreasing late work at lower remote frequencies ($<$ 1.5 days/week) positively mediates the promotion effect, while increasing late work at high remote frequencies ($\geq$ 4 days/week) results in a negative mediation effect.