Harnessing mobility data to capture changing work from home behaviours between censuses
Hamish Gibbs; Patrick Ballantyne; James Cheshire; Alex Singleton; Mark A. Green (2024). The Geographical Journal, 190(2). DOI: 10.1111/geoj.12555
Abstract
This paper provides an analysis of working from home patterns in England using data from the 2021 Census to understand (1) how patterns of working from home (WFH) in England have shifted since the COVID‐19 pandemic and (2) whether human mobility indicators, specifically Google Community Mobility Reports, provide a reliable proxy for WFH patterns recorded by the 2021 Census, providing a formal evaluation of the reliability of such datasets, whose applications have grown exponentially over the COVID‐19 pandemic. We find that WFH patterns recorded by the 2021 Census were unique compared with previous UK censuses, reflecting an unprecedented increase likely caused by persistent changes to employment during the COVID‐19 pandemic, with a clear social gradient emerging across the country. We also find that Google mobility in ‘Residential’ and ‘Workplace’ settings provides a reliable measurement of the distribution of WFH populations across Local Authorities, with varying uncertainties for mobility indicators collected in different settings. These findings provide insights into the utility of such datasets to support population research in intercensal periods, where shifts may be occurring, but can be difficult to quantify empirically.
Extended Summary
This research examines how working from home patterns in England changed during the COVID-19 pandemic and whether Google mobility data can accurately measure these changes. The study analysed three decades of UK census data (2001, 2011, and 2021) alongside Google Community Mobility Reports covering the pandemic period from February 2020 to October 2022. The research focused on Local Authority District level data to understand geographical variations in remote working behaviours. The findings reveal that working from home increased dramatically between 2011 and 2021, rising from an average of 5.3% to 14.7% of the population across English local authorities. This represents an unprecedented shift compared to the relatively stable patterns observed between 2001 and 2011. The research identifies a clear socioeconomic gradient in working from home adoption, with less deprived areas experiencing significantly higher increases than more deprived regions. London emerged as a particular case study, showing the highest working from home rates regardless of local deprivation levels, likely reflecting the concentration of knowledge-based industries that could more easily adapt to remote working arrangements. The study demonstrates that Google mobility data, particularly indicators measuring time spent in residential areas and visits to workplace locations, provides a reliable proxy for census-recorded working from home patterns. The correlation between Google’s residential mobility data and census working from home figures reached 0.93, whilst workplace mobility showed a correlation of 0.86. This strong relationship suggests that commercial mobility datasets could potentially supplement official statistics during intercensal periods. However, the research also reveals important limitations in using mobility data for predictive purposes. When attempting to forecast working from home patterns beyond the census date, the models produced implausible results, suggesting that whilst mobility data captures patterns at specific time points, temporal variations and seasonal effects limit its utility for long-term projection. The study’s broader implications extend beyond methodological considerations to urban planning and policy development. The uneven geographical distribution of working from home adoption may reinforce existing regional inequalities, with economic activity shifting from central business districts to residential areas in affluent neighbourhoods. This spatial reorganisation of work patterns requires adaptive planning strategies for transport infrastructure, office space utilisation, and local economic development. The research contributes to ongoing debates about the future role of official statistics, particularly as the UK considers modernising its census methodology. Whilst the study demonstrates that alternative data sources like mobility indicators can provide valuable insights into population behaviours, they cannot yet replace the comprehensive and standardised nature of census data.
Key Findings
- Working from home in England increased from 5.3% in 2011 to 14.7% in 2021, representing unprecedented pandemic-driven change
- Google mobility data shows strong correlation (0.93) with census working from home patterns, validating commercial datasets for population research
- Clear socioeconomic gradient emerged with least deprived areas experiencing 12.8% increase versus 8.2% in most deprived areas
- London displayed unique patterns with high working from home rates regardless of local deprivation levels due to industry composition
- Mobility data proves reliable for measuring current patterns but unsuitable for long-term forecasting due to temporal variations
Citation
@article{gibbs2024harnessing,
author = {Hamish Gibbs; Patrick Ballantyne; James Cheshire; Alex Singleton; Mark A. Green},
title = {Harnessing mobility data to capture changing work from home behaviours between censuses},
journal = {The Geographical Journal},
year = {2024},
volume = {190(2)},
doi = {10.1111/geoj.12555}
}