Source-linked AI summary
Pandemic Programming: How COVID-19 affects software developers and how their organizations can help
Paul Ralph, Sebastian Baltes, Gianisa Adisaputri, Richard Torkar, Vladimir Kovalenko, Marcos Kalinowski, Nicole Novielli, Shin Yoo, Xavier Devroey, Xin Tan, Minghui Zhou, Burak Turhan, Rashina Hoda, Hideaki Hata, Gregorio Robles, Amin Milani Fard, Rana Alkadhi
TL;DR
The study examines how crisis-driven home working affected software developers’ wellbeing and productivity, an area with limited evidence from pandemic conditions. Using a multilingual questionnaire based mainly on validated scales and structural equation modeling, it finds declines in both outcomes, close interrelation, and roles for preparedness, fear, and home-office ergonomics.
Problem
The study addresses limited evidence about how pandemic home working affects software developers’ wellbeing and productivity and how organizations can support them.
Method
A multilingual questionnaire using mainly validated scales was analyzed with factor analysis, non-parametric inferential statistics, and structural equation modeling.
Results
The pandemic was associated with diminished developer wellbeing and productivity, which were closely related; preparedness, fear, and home-office ergonomics also affected these outcomes.
Takeaways & Limitations
Software companies should prioritize employee wellbeing and home-office ergonomics, while providing differentiated support because women, parents, and people with disabilities may be disproportionately affected.
Takeaways & Limitations
Convenience and snowball sampling may bias the sample in unknown ways, limiting assessment of representativeness and generalizability.
Abstract
from arXiv · showhide
Context. As a novel coronavirus swept the world in early 2020, thousands of software developers began working from home. Many did so on short notice, under difficult and stressful conditions. Objective. This study investigates the effects of the pandemic on developers' wellbeing and productivity. Method. A questionnaire survey was created mainly from existing, validated scales and translated into 12 languages. The data was analyzed using non-parametric inferential statistics and structural equation modeling. Results. The questionnaire received 2225 usable responses from 53 countries. Factor analysis supported the validity of the scales and the structural model achieved a good fit (CFI = 0.961, RMSEA = 0.051, SRMR = 0.067). Confirmatory results include: (1) the pandemic has had a negative effect on developers' wellbeing and productivity; (2) productivity and wellbeing are closely related; (3) disaster preparedness, fear related to the pandemic and home office ergonomics all affect wellbeing or productivity. Exploratory analysis suggests that: (1) women, parents and people with disabilities may be disproportionately affected; (2) different people need different kinds of support. Conclusions. To improve employee productivity, software companies should focus on maximizing employee wellbeing and improving the ergonomics of employees' home offices. Women, parents and disabled persons may require extra support.
1 Introduction
The pandemic created an unusual and difficult work-from-home context, leaving software companies with limited evidence about how to support developers. The study therefore asks how pandemic-related home working affects developers’ wellbeing and productivity and develops a model to explain these changes.
- Working from home during the pandemic differed from normal remote work because people faced improvised workspaces, distractions, isolation, childcare, and home-schooling demands.
- Few studies had examined working from home during disasters, and none had studied a pandemic of this magnitude in the modern internet era.
- The research question asks how working from home during COVID-19 affected software developers’ emotional wellbeing and productivity.
- The study generates and evaluates a theoretical model explaining and predicting changes in wellbeing and productivity during crisis-related home working.
2 Background
Prior research shows that disasters disrupt work and that remote work can affect productivity and wellbeing differently across people and conditions. Pandemic home working is especially constrained by inadequate organizational preparation, caregiving demands, isolation, and uncertain productivity measurement.
- Disasters can reduce work quality and productive capacity, creating a need for employer strategies that support employees and business continuation.
- Pandemic productivity effects may vary by person, project, and metric, with some evidence of longer working hours at an unsustainable pace.
- Pandemic home working differs from ordinary telework because organizations and workers may lack dedicated workspace, supportive policies, resources, and management practices.
- Remote work is often associated with higher perceived productivity, but organizational and job-related factors can strongly affect satisfaction and perceived productivity.
- Wellbeing effects of remote work vary with emotional stability: autonomy may foster wellbeing for some people, while remote work can increase strain for others.
- Perceived productivity is widely used but remains contested because self-reports may inflate benefits and simple output metrics have low construct validity.
3 Hypotheses
The hypotheses predict declines in wellbeing and perceived productivity during pandemic home working, a direct relationship between them, and effects from preparedness, fear, and home-office ergonomics.
- The study hypothesizes that developers will experience lower wellbeing and lower perceived productivity after switching home because of COVID-19.
- The proposed model treats changes in wellbeing and perceived productivity as related constructs that must change after the transition to home working.
- Change in wellbeing and change in perceived productivity are hypothesized to be directly related, potentially forming a reciprocal downward spiral.
- Disaster preparedness is hypothesized to relate directly to changes in wellbeing and perceived productivity, with lack of preparedness expected to exacerbate reductions.
- Fear of the pandemic is hypothesized to be inversely related to changes in wellbeing and perceived productivity.
- Home-office ergonomics is hypothesized to relate directly to changes in wellbeing and perceived productivity, with safer and more comfortable environments expected to support both.
4 Method
The study used a multilingual anonymous questionnaire for software professionals who switched from office work to home working because of COVID-19. It relied mainly on validated scales while developing measures for ergonomics and organizational crisis support.
- The questionnaire was translated and localized into 12 languages, with region-specific advertising strategies used during data collection.
- The target population was software developers worldwide who switched from working in an office to working from home because of COVID-19.
- The anonymous survey included software professionals but primarily targeted developers and collected demographic, household, and pandemic-related information.
- Validated scales were used where possible to improve construct validity for latent concepts such as fear, preparedness, ergonomics, wellbeing, and productivity.
- Wellbeing was measured with the WHO-5 before and after working from home, while perceived productivity was measured with Health and Work Performance Questionnaire items.
- Ergonomics was measured with a six-item, six-point Likert scale covering distractions, noise, lighting, temperature, chair comfort, and overall ergonomics.
- Organizational support was developed from interviews and literature into 22 actions across equipment, reassurance, connectedness, self-care, and technical infrastructure or practices.
- Convenience and snowball sampling limited the ability to evaluate sample representativeness and generalizability accurately.
5 Analysis and Results
The analysis retained 2225 responses after cleaning and found minimal response bias, while characterizing participants across demographics, living arrangements, work experience, countries, and organization sizes.
- Data cleaning: 2225 usable responses remained after excluding 439 responses that failed inclusion criteria and 4 effectively blank responses.The cleaning process also removed identifying or redundant fields and added a binary field.
- Validity analysis: ∆Productivity 7 and 9 were dropped because of possible loading issues, while Ergonomics 1 was retained after loadings stabilized.Dropping the two productivity indicators resolved the issue affecting Ergonomics 1.
- Response bias: The response-bias analysis found significant differences only for adult cohabitants and age, with very small effect sizes, consistent with minimal response bias.The study could not compare the sample with known population parameters or use conventional early-versus-late respondent comparisons.
- Participant demographics: Participants came from 53 countries and organizations ranging from 0–9 employees to more than 100,000; 80% identified as software developers or equivalent.Mean work experience was 9.3 years, while 58% had no prior experience working from home.
5.4 Change in wellbeing and productivity
Developers reported lower wellbeing and perceived productivity after switching to pandemic-era home working. Structural equation modeling supported relationships among wellbeing, productivity, ergonomics, disaster preparedness, and fear, with acceptable overall fit.
- Wilcoxon signed-rank tests supported lower wellbeing (V = 645610; p < 0.001; δ = 0.12 ± 0.03) while working from home.Wellbeing was measured before and after the switch using the WHO5 scale.
- Wilcoxon signed-rank tests supported lower perceived productivity (V = 566520; p < 0.001; δ = 0.13 ± 0.03) while working from home.Productivity was measured before and after the switch using the HPQ scale.
- The structural model used SEM to estimate relationships among latent constructs and directly measured control variables.Confirmatory factor analysis mapped reflective indicators to constructs before structural regressions were estimated.
- CFI = 0.961, RMSEA = 0.051, and SRMR = 0.067 indicated that the structural model was safe to interpret.The model converged after 97 iterations with 212 free parameters and n = 1377.
- Changes in wellbeing and perceived productivity were directly related, while ergonomics, disaster preparedness, and fear were linked to changes in wellbeing or productivity.Disaster preparedness was inversely related to fear; supported hypotheses included H1–H3, H5, H6, and H8–H10.
5.6 Exploratory findings
Exploratory analyses identified demographic and situational differences in preparedness, fear, home-office ergonomics, wellbeing, and productivity. Some effects were conflicting or difficult to interpret, motivating further research.
- People with small children had less ergonomic home offices, while women tended to be more fearful and people with disabilities were less prepared and less ergonomically equipped.The reported patterns concern noise, distractions, fear, disaster preparedness, and home-office ergonomics.
- People with COVID-19 exposure tended to be more afraid, more prepared, and to have worse wellbeing after switching to home working.Exposure included having COVID-19 or having affected family members, housemates, or close friends.
- People living with other adults were more prepared, people living alone had more ergonomic home offices, and more isolated people tended to be more afraid.Isolation was defined as not leaving home at all or leaving only for necessities.
- Changes in productivity and wellbeing were closely related, with possible mediation making some indirect effects difficult to interpret.The authors note that productivity may mediate disaster preparedness’s relationship with wellbeing, while wellbeing may mediate fear’s relationship with productivity.
- Disability showed conflicting effects, with a direct positive effect on productivity and an indirect negative effect through fear.The authors state that more research is needed to explore these relationships.
- Country and language were significant predictors of the latent variables in an attempted model, but including many binary indicators made the model impossible to interpret.The authors therefore call for more research on the nature of country, language, and likely cultural effects.
5.7 Organizational support
Participants did not agree broadly on which organizational supports would help them. Paying home internet charges was the only listed action viewed as helpful by more than half of participants, while regular meetings were widely viewed as unhelpful.
- Paying developers’ home internet charges was the only action perceived as helpful by more than half of participants.Less than 10% of companies appeared to be providing this support.
- Regular meetings were the action most companies were taking, but most participants did not perceive them as helpful.The passage contrasts organizational practice with participant-perceived helpfulness.
- There was no apparent correlation between supports developers believed would help and actions employers were actually taking.The finding concerns the relationship between perceived helpfulness and reported organizational behavior.
- Developers showed little consensus about what their organizations should do to help them.The authors suggest that different participants may value different forms of support.
- The support-action question may have reduced discrimination between items, and helpfulness may vary by country and social safety net.The authors propose asking participants for their top N actions in future work to improve reliability.
5.8 Summary interpretation
The study reports diminished wellbeing and productivity among software professionals working from home during the pandemic, with the two outcomes closely related. It also identifies preparedness, fear, ergonomics, and demographic differences as relevant factors.
- Software professionals working from home during the pandemic experienced diminished emotional wellbeing and productivity, which were closely related.
- Poor disaster preparedness, pandemic-related fear, and poor home-office ergonomics were associated with reduced wellbeing and productivity.
- Women, parents, and people with disabilities may be disproportionately affected, while organizational support preferences showed dissensus.
6 Discussion
The discussion recommends supporting wellbeing rather than demanding normal productivity, while recognizing that support needs vary and that the study has important sampling, measurement, and causal-validity limitations.
- Recommendations: Organizations should avoid judging employees by pandemic productivity because such decisions may disadvantage protected groups.The discussion specifically warns against using productivity to guide layoffs, promotions, or bonuses during the pandemic.
- Recommendations: Because wellbeing and productivity are closely related, improving employee wellbeing is presented as the best route to improving productivity.The authors recommend discussing individual needs with employees because no single organizational action benefits everyone.
- Recommendations: Improving home-office ergonomics may help, but organizations should address broad comfort and distraction needs rather than micromanage physical positioning.Examples include providing an office chair or noise-cancelling headphones.
- Limitations: The study’s convenience and snowball sampling may bias the sample in unknown ways, despite localized multilingual recruitment.The authors recruited through local co-authors who translated, localized, and advertised the questionnaire.
- Limitations: Perceived productivity may not represent actual productivity, and the relationship between reported organizational actions and helpfulness remains uncertain.The authors note that the productivity scale may not correlate with objective software-development performance during pandemics.
- Limitations: Structural equation modeling supports the model’s fit but cannot establish temporal precedence, and several third-variable explanations remain possible.Potential confounders include overtime, personality, team dynamics, organizational culture, family conflict, medical history, and wealth.
7 Conclusion
The conclusion presents the pandemic as a distinct working-from-home context for software developers and summarizes evidence of reduced wellbeing and productivity, varied support needs, and disproportionate effects on some groups.
- Conclusion: Pandemic working-from-home conditions differ fundamentally from normal remote work because developers face stress, isolation, restrictions, closures, and disrupted services.The conclusion identifies the study as a large-scale investigation of these conditions among software developers.
- Conclusion: The study reports declines in productivity and wellbeing, a close relationship between them, and a model explaining and predicting pandemic effects.These are listed as central contributions of the study.
- Conclusion: Different people may need different organizational support, with no single universally effective intervention.The conclusion describes this as an indication rather than a universal rule.
- Conclusion: The pandemic may disproportionately affect women, parents, and people with disabilities.The conclusion presents this as an indication requiring further attention.
- Conclusion: The study combines validated and re-validated scales, 2225 responses, 12 languages, structural equation modeling, and disaster-management perspectives.These features are presented as distinguishing characteristics of the study.
- Conclusion: The authors call for more research on software development during crises, pandemics, lockdowns, and other adverse conditions.They specifically connect this need to future work spanning disaster management and software engineering.
8 Data Availability
The paper provides an anonymous dataset, instruments, and analysis scripts through an open-data archive, while the acknowledgements identify supporting institutions and grants.
- Data Availability: A comprehensive replication package containing the anonymous dataset, instruments, and analysis scripts is available in the Zenodo open-data archive.The paper gives the archive record URL for access.
- Acknowledgements: The project received support from research councils, governments, and universities, and acknowledges additional advisers and media outlets.The listed support includes Canadian, Spanish, Dalhousie University, and University of Adelaide funding or institutional support.