Interesting initiative, but I see many issues in the study:
- As others have mentioned, it is not generalizable to other types of jobs. Call-center work may be improved when working from home, but what about other types of jobs requiring more interaction/communication
- But the biggest issue I see is with their mention of a "Randomized experiment". Well, technically it is randomized among those who volunteered for the experiment. So from a causal inference perspective, the results generalize only to the population of employees who volunteered, not to the entire company.
More generally, one of the problems with these types of study is that the best it can give is an average gain from home working. But suppose the gain is null. It doesn't mean that this is not a viable option. Maybe there is a huge variance in the gain, depending on the individuals: maybe the gain in productivity is huge for some employees, while it is negative for some others. This could depend on the employee's self-discipline (whatever that means), his family situation (single/married, how many kids), where he leaves, is he introvert, etc... A useful study would be to randomized the assignment, stratifying by according to these covariates, in order to learn more about their effect on the performance of the employees.
- As others have mentioned, it is not generalizable to other types of jobs. Call-center work may be improved when working from home, but what about other types of jobs requiring more interaction/communication
- But the biggest issue I see is with their mention of a "Randomized experiment". Well, technically it is randomized among those who volunteered for the experiment. So from a causal inference perspective, the results generalize only to the population of employees who volunteered, not to the entire company.
More generally, one of the problems with these types of study is that the best it can give is an average gain from home working. But suppose the gain is null. It doesn't mean that this is not a viable option. Maybe there is a huge variance in the gain, depending on the individuals: maybe the gain in productivity is huge for some employees, while it is negative for some others. This could depend on the employee's self-discipline (whatever that means), his family situation (single/married, how many kids), where he leaves, is he introvert, etc... A useful study would be to randomized the assignment, stratifying by according to these covariates, in order to learn more about their effect on the performance of the employees.