From Gen Z, Millennials, to Babyboomers: Portraits of Working from Home during the COVID-19 Pandemic

Ziyu Xiong, Pin Li, Hanjia Lyu, Jiebo LuoarXiv 2021 · arXiv

Abstract

This study provides a large-scale demographic analysis of the transition to remote work during the early COVID-19 pandemic. By leveraging machine learning on Twitter data from over 28,000 users, researchers from the University of Rochester quantified public sentiment and identified key socioeconomic drivers. The findings reveal that while the overall outlook on working from home (WFH) was positive, sentiment was significantly higher among women, individuals in their 40s, and residents of high-income urban areas.

Introduction

The COVID-19 lockdowns of early 2020 forced a massive, overnight shift to remote work for millions. Traditional surveys often struggle with small sample sizes and temporal limitations. This research addresses these gaps by using social media as a real-time sensor for public sentiment. The authors sought to identify which demographic characteristics correlate with WFH experiences and what specific topics—such as family, policy, or productivity—shaped those feelings.

Methodology

The researchers employed a multi-stage computational pipeline to turn unstructured social media data into a structured demographic dataset:

  • Sentiment Analysis: Used VADER (Valence Aware Dictionary and sEntiment Reasoner) to calculate sentiment scores ranging from -1 (negative) to +1 (positive).
  • Demographic Inference: Applied the M3-inference deep learning model to predict age and gender based on profile pictures, usernames, and bios.
  • Socioeconomic Mapping: Inferred ethnicity via the Ethnicolr API and mapped median household income and population density using 2019 U.S. Census (ACS) data.
  • Topic Modeling: Used Latent Dirichlet Allocation (LDA) to categorize tweets into nine distinct themes, such as "Family activities," "Government and policy," and "Back to office."

The final analysis focused on 28,579 unique Twitter users collected between April 5 and April 26, 2020.

Results and Discussion

Sentiment by Topic

The average sentiment score across the dataset was 0.242, indicating a generally favorable view of WFH. However, sentiment varied widely by topic.

Average sentiment scores across nine LDA topics, showing Topic 7 (Encouragement) as the most positive and Topic 5 (Government and policy) as the least.

As illustrated in the figure above, "Encouragement" (Topic 7) was the most positive theme (0.460), while "Government and policy" (Topic 5) recorded the lowest sentiment (0.129).

Demographic Insights

The study identified several statistically significant predictors of positive WFH sentiment through OLS regression:

  • The Gender Gap: Women expressed significantly more positive sentiment than men ($p < .001$). Researchers suggest WFH offered flexibility to manage professional work alongside childcare, despite the added stress of the "double burden."
  • The 40s Sweet Spot: Contrary to expectations that younger "digital natives" would be most enthusiastic, individuals in their 40s showed the highest positive sentiment ($p < .001$). This group likely prioritized the safety of WFH and the stability it provided for retirement savings during market volatility.
  • Income Divide: Sentiment strongly correlated with income ($p < .001$). High-income individuals, typically in "knowledge work" roles, found the transition easier than those in lower-income areas.

Permeable Boundaries

Topic modeling highlighted how the focus of WFH discussions evolved with age. Younger users (0-18) were most likely to tweet about "Family activities" (23.4%), while older users (40+) shifted their focus toward the logistics of remote work.

Topic distribution for users aged 40 and above, highlighting a significant focus on "Remote" work/study topics (23.5%).

As shown in the pie chart for the 40+ demographic, "Remote" work and study topics accounted for nearly a quarter of all discussions, reflecting a high degree of professional integration in the home environment compared to younger cohorts.

Conclusion

This research underscores that the WFH experience is not monolithic. Demographic factors like age, gender, and income create distinct "portraits" of remote work. While the study is limited by platform-specific demographics and the lack of occupation data, it provides a valuable framework for understanding how different segments of society adapt to major labor shifts. Future research could expand these findings by incorporating professional backgrounds and longitudinal data to track sentiment evolution beyond the initial lockdown phase.

Summary updated

Abstract

Since March 2020, companies nationwide have started work from home (WFH) due to the rapid increase of confirmed COVID-19 cases in an attempt to help prevent the coronavirus from spreading and rescue the economy from the pandemic. Many organizations have conducted surveys to understand people's opinions towards WFH. However, the findings are limited due to small sample size and the dynamic topics over time. This study aims to understand the U.S. public opinions on working from home during the COVID-19 pandemic. We conduct a large-scale social media study using Twitter data to portrait different groups who have positive/negative opinions about WFH. We perform an ordinary least squares regression to investigate the relationship between the sentiment about WFH and user characteristics including gender, age, ethnicity, median household income, and population density. To better understand public opinion, we use latent Dirichlet allocation to extract topics and discover how tweet contents relate to people's attitudes. These findings provide evidence that sentiment about WFH varies across user characteristics. Furthermore, the content analysis sheds light on the nuanced differences in sentiment and reveals disparities relate to WFH.
BibTeX
@misc{xiong2021from,
  author = {Ziyu Xiong and Pin Li and Hanjia Lyu and Jiebo Luo},
  title = {{From Gen Z, Millennials, to Babyboomers: Portraits of Working from Home during the COVID-19 Pandemic}},
  year = {2021},
  eprint = {2101.06762},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2101.06762},
}
Figure