��ɫ��Ӱ

We surveyed college graduates in 40 countries across the Americas, Europe, Asia and Africa during November 2024 to February 2025 to collect information on working from home (WFH). We find five facts. (1) WFH is highest in North America, UK and Australia, and lowest in Asia; (2) WFH levels fell from 2022 to 2023 but have since stabilized; (3) employees with children are more likely to split their workweeks between home and employer’s location, while those without children are more likely to work in a fully remote or fully onsite capacity; (4) WFH levels are similar for men and women in every major region of the world; and (5) the desire to WFH is highest among women with children.

Our latest Global Survey of Working Arrangements (G-SWA) collected data from over 16,000 college and university graduates across 40 countries from November 2024 through February 2025.¹ We targeted graduates as they are more likely to work in positions that have the potential for some work from home (WFH). Analyzing our data yields five key results.

1) WFH is highest in North America and Europe and lowest in Asia

Figure 1 plots the level of working from home across countries. English-speaking countries have on average the highest levels of WFH at about 1.5 to 2 days a week, European countries are a little below at around 1 to 1.5 days a week, Latin American and African countries are lower still at around 1 day per week, and Asian countries are the lowest at 0.5 to 1 day per week. This global working from home ranking is very similar to the results for our 2023 survey wave (see Aksoy et al., 2023b).

2) WFH levels have stabilized since 2023

Figure 2 plots the average levels of WFH for our last three survey waves. Overall levels of working from home in our global survey declined from an average of 1.6 days in 2022 to 1.33 days in 2023 to 1.27 days in 2024/2025. This is calculated for a balanced panel of 22 countries which we surveyed in each of the three waves. These results highlight how working from home levels dropped from 2022 to 2023, but after 2023 appear to have stabilized. This mimics the pattern in US data, as highlighted in Buckman et al. (2025).

3) Hybrid working arrangements are more common for parents

Figure 3 plots WFH rates for men and women with and without children. The left panel reveals little difference in overall WFH rates between persons with children and those without children. This pattern holds for men and women.  

The right panel reveals a different story when we distinguish fully remote work from hybrid arrangements, whereby someone works some days at home and other days at the employer’s (or client’s) worksite. In particular, employees with children are much more likely to have a hybrid arrangement that involves 1, 2 or 3 days a week of working from home. Employees without children are more likely to be fully in person with 0 days a week at home, or to be fully remote with 5+ day a week at home.

4) WFH rates are similar for men and women

Figure 4 shows WFH rates are similar for men and women across all major regions covered by our survey. We see some limited differences across regions, but the overall pattern is that men and women have similar WFH rates within countries.

5) Desires to WFH are highest for women with children

While men and women have similar WFH rates, women with children have slightly stronger desires to work from than women without children. Women with children desire to WFH an average of 2.66 days a week, 0.13 days more than women without children. In contrast for men, those with and without children have very similar desires, and indeed men without children have a slightly higher WFH preference.

Appendix

The fourth wave of the G-SWA has been fielded in 40 countries from November 2024 to February 2025 (see Appendix Table 1). The survey includes two equally sized subsamples in each country. The first subsample consists of adults aged 20 to 64, while the second is restricted to full- time working adults aged 20 to 64 who have completed at least secondary education. In France, Germany, Italy, the UK and the US, total sample sizes amount to more than 2,500 respondents, respectively. In all other countries, total samples consist of roughly 1,000 responses.² In this report, we restrict attention to full-time workers, aged 20-64, with completed tertiary education, coming from both subsamples. 

In addition to basic questions on demographics, employment status, earnings, industry, occupation, marital status and living arrangements, the survey asks about current, planned and desired WFH levels, and more. We screen out respondents who fail to answer an attention check at the beginning of the survey.³ We design the G-SWA instrument, adapting questions from the U.S. SWAA developed by Barrero et al. (2021). We enlist professionals to translate our original English-language questionnaire into the major languages of each country. To ensure high-quality translations, we also enlist an independent third party with knowledge of the survey to review the translations and revise as needed. 

To field the G-SWA, we contract with (a professional survey firm), which implements the survey directly and in cooperation with its external partners. The survey effort taps pre-recruited panels of people who previously expressed a willingness to take part in research.⁴ Recruitment into these panels happens via partner affiliate networks, multiple advertising channels (including Facebook, Google Adwords, and other websites), address databases, and referrals. New recruits are added to the panel on a regular basis. When it is time to field a survey, Bilendi or its partner issues email messages that invite panel members to participate. The message contains information about compensation and estimated completion time but not about the survey topic. Clicking on the link in the invitation message takes the recipient to the online questionnaire. Respondents who complete the survey receive cash, vouchers or award points, which they can also donate.⁵ 

Before our analysis of the G-SWA data, we drop “speeders,” defined as respondents in the bottom 5% of the completion-time distribution for each country. Additionally, we remove those who fail a second attention check question.⁶ After these drops, our analysis sample contains 16,422 observations across the 40 countries in Wave 4. Appendix Table A.1 reports statistics on response time, observation counts and dates in the field for each country. Our samples are broadly representative by age, gender, and education for the group of full-time workers in each country.⁷

References

Aksoy, Cevat Giray, Jose Maria Barrero, Nicholas Bloom, Steven J. Davis, Mathias Dolls, and Pablo Zarate. 2022. “”, Brookings Papers on Economic Activity 53 (2): 281-330.

Aksoy, Cevat Giray, Jose Maria Barrero, Nicholas Bloom, Steven J. Davis, Mathias Dolls, and Pablo Zarate. 2023a. “”, AEA Papers and Proceedings, Vol. 113, May 2023, 597-603.

Aksoy, Cevat Giray, Jose Maria Barrero, Nicholas Bloom, Steven J. Davis, Mathias Dolls, and Pablo Zarate. 2023b. “.” EconPol Policy Brief 53.

Barrero, Jose Maria, Nicholas Bloom, and Steven J. Davis. 2021. “,” NBER Working Paper 28731.

Buckman, Shelby, Jose Maria Barrero, Nicholas Bloom, and Steven J. Davis. 2025. “,” NBER Working Paper 33508.

Table A.1: Statistics on Response Time (in minutes), Sample Size, and Dates in the Field

Figure 1: Working from home is more common in North America and Europe, and less common in Asia, College-Educated Workers

Note: Responses to the question “For each day last week, did you work 6 or more hours, and if so where?” N=16,422 college-educated workers in 40 countries surveyed in November 2024 – February 2025.

Figure 2: Work from home levels have stabilized since 2023

Note: Responses to the question “For each day last week, did you work 6 or more hours, and if so where?”. N=40,751 college-educated workers in 22 countries surveyed in 2022, 2023 and 2024.

Figure 3: Hybrid work schedules are more common for parents

Note: Responses to the question “For each day last week, did you work 6 or more hours, and if so where?” N= 16,422 college-educated workers in 40 countries surveyed in November 2024 – February 2025.

Figure 4: College-Educated Men and women have similar WFH levels

Note: Responses to the question “For each day last week, did you work 6 or more hours, and if so where?”. N= 16,422 college-educated workers in 40 countries surveyed in November 2024 – February 2025.

Figure 5: Women with children have a stronger desire to work from home

Note: Responses to the question “Looking one year ahead, how often would you like to have paid workdays at home?” N= 16,422 college-educated workers in 40 countries surveyed in November 2024 – February 2025.

Acknowledgements

We gratefully acknowledge financial support from the Asian Development Bank and the Japan Funds for Prosperity and Resilience Asia and the Pacific, German Federal Ministry of Education and Research through the project “Development of poverty and inequality in Germany and France over the past 25 years” (grant number 01UI2208A), EBRD, King’s College London, Smith Richardson Foundation, and World Bank. The opinions and conclusions expressed herein are solely those of the authors and do not reflect the opinions of any sponsoring agency.

Authors

• Cevat Giray Aksoy. European Bank for Reconstruction and Development and King’s College London cevat.aksoy@kcl.ac.uk
• Jose Maria Barrero. Instituto Tecnológico Autónomo de México jose.barrero@itam.mx
• Nicholas Bloom. ��ɫ��Ӱ nbloom@stanford.edu
• Steven J. Davis. Hoover Institution at ��ɫ��Ӱ StevenD5@��ɫ��Ӱ.edu
• Mathias Dolls. ifo Institute dolls@ifo.de
• Pablo Zarate. Princeton University pzarate@princeton.edu 

Footnotes

1. Aksoy et al. (2022, 2023 a, b) report results from the previous three G-SWA waves, conducted in July-August 2021 (1st wave), January-February 2022 (2nd wave) and April-May 2023 (3rd wave).
2. The sample size in India and Nigeria is somewhat smaller and amounts to 875 respondents.
3. The attention check reads: “What is 3+4?” 4) Bilendi and its external partners do not engage in “river sampling,” whereby people are invited to take a survey while engaging in another online activity. Relative to river sampling, the use of pre-recruited panels affords greater control over sample composition and selection.
4. Bilendi and its external partners do not engage in “river sampling,” whereby people are invited to take a survey while engaging in another online activity. Relative to river sampling, the use of pre-recruited panels affords greater control over sample composition and selection.
5. We do not contact respondents ourselves, do not collect personally identifiable information, and have no way to re-contact them.
6. The attention check reads: “In how many big cities with more than 500.000 inhabitants have you lived? Irrespective of the truth, please insert the number 33 in order to continue with the survey”.
7. Respondents take the survey on a computer, smart-phone, iPad or like device, so we miss persons who don’t use such devices.