How many remote workers, and under which definition

A survey may count people who worked remotely at least once in the reference week, at least once in the month, or regularly above a threshold number of days.

It may also count people whose job could be done remotely, which is an estimate of feasibility rather than an observed practice. These four measures produce four different figures.

The legal definition adds another layer. The French Labour Code covers work that could have been carried out on the employer's premises and is done elsewhere on a voluntary basis, using information technology. Someone who finishes a file at home in the evening does not always fall within that definition, but may answer yes to a survey.

Surveys changed their questions from 2020 onwards to capture a phenomenon that had changed in scale. These changes create breaks in the series that statistical institutes flag, and that commentary often ignores.

Comparing a pre-2020 figure with a later one therefore requires checking that the instrument has not changed, which is rarely the case. Even the wording of a question, or the example given to the respondent, is enough to shift the answers of people whose practice is occasional.

The rise compared with the earlier period is robust in every country with data series, whatever the definition used, and the level has stabilised after falling back from the peak.

Remote work remains very unevenly distributed: it is concentrated in managerial and professional occupations and in service activities, and its feasibility depends on job characteristics that occupational classifications capture imperfectly. The dominant form is hybrid, a few days a week, rather than fully remote work.

Estimates of the share of jobs that can be done remotely rest on task classifications by occupation, built from descriptive databases of occupations. The most cited, published in 2020 by Jonathan Dingel and Brent Neiman, draws on an American database describing occupations.

These estimates vary widely with the method and proved imprecise once comparison with observed practice became possible, which suggests treating them as orders of magnitude.

They measure what the tasks allow, not what the employer accepts or what the person wants. Two technically identical jobs can see opposite practices depending on the organisation, something no occupational database can anticipate.

Surveys count days, not conditions. They say little about the space available at home, about sharing it with other occupants or about who bears the costs.

These dimensions appear in a few qualitative surveys and in company-level bargaining, but they do not exist in national series, which limits what can be said about the actual conditions of the work.