Working Paper: NBER ID: w31515
Authors: David Atkin; Antoinette Schoar; Sumit Shinde
Abstract: Differences in the productivity of home-based versus office-based work may arise due to a treatment effect of the office or from workers with different abilities sorting into these locations. We conduct an RCT in the data entry sector in India that exogenously allocates workers to work from home (WFH) or from the office. We first find that the productivity of workers randomly assigned to WFH is 18% lower than those in the office. Two-thirds of the effect manifests itself from the first day of work with the remainder due to quicker learning in the office. Second, there is a negative selection effect into the office, since workers who prefer home-based work are 12% faster and more accurate at baseline. Finally, we find negative selection on treatment effects: workers who prefer WFH are substantially less productive at home than the office (27% less compared to 13% less for workers who prefer the office). These negative selection effects are partially explained by subgroups that face bigger constraints on selecting into office work and additional demands on their attention when at home, such as those with children or other home care responsibilities.
Keywords: productivity; remote work; RCT; worker sorting; India
JEL Codes: J01; J10; J22; J24; J46
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| work location (home) (J29) | productivity (O49) |
| work location (office) (J29) | productivity (O49) |
| work location (home) (J29) | slower learning (J24) |
| preference for WFH (J29) | productivity when working from home (J29) |
| preference for office work (J29) | productivity when working from home (J29) |
| family care responsibilities (J13) | productivity outcomes (O49) |
| older female workers (J21) | productivity losses when working from home (J22) |
| workers with children (J82) | productivity losses when working from home (J22) |