Source-linked AI summary
Predictors of Well-being and Productivity among Software Professionals during the COVID-19 Pandemic -- A Longitudinal Study
Daniel Russo, Paul H. P. Hanel, Seraphina Altnickel, Niels van Berkel
TL;DR
The paper examines which factors predict software engineers’ well-being and productivity while working remotely during pandemic restrictions. Using a two-wave longitudinal study of 192 software professionals, it finds consistent associations for social contacts, stress, boredom, and distractions, but no evidence of causal effects.
Problem
The study addresses limited evidence on how pandemic-related remote work affects software engineers’ well-being and productivity.
Method
A two-wave longitudinal study of 192 software professionals examined over 50 candidate factors using correlations, regressions, and structural equation modeling.
Results
Quality of social contacts positively predicted well-being, stress negatively predicted well-being, and boredom and distractions negatively predicted productivity, while longitudinal analyses found no significant causal relations.
Takeaways & Limitations
Working from home was not, by itself, a significant challenge for software engineers, although stress, social contacts, boredom, and distractions were associated with outcomes.
Takeaways & Limitations
The sample lacked sufficient participants from different countries to test whether government lockdown strictness moderated the observed associations.
Abstract
from arXiv · showhide
The COVID-19 pandemic has forced governments worldwide to impose movement restrictions on their citizens. Although critical to reducing the virus' reproduction rate, these restrictions come with far-reaching social and economic consequences. In this paper, we investigate the impact of these restrictions on an individual level among software engineers who were working from home. Although software professionals are accustomed to working with digital tools, but not all of them remotely, in their day-to-day work, the abrupt and enforced work-from-home context has resulted in an unprecedented scenario for the software engineering community. In a two-wave longitudinal study (N=192), we covered over 50 psychological, social, situational, and physiological factors that have previously been associated with well-being or productivity. Examples include anxiety, distractions, coping strategies, psychological and physical needs, office set-up, stress, and work motivation. This design allowed us to identify the variables that explained unique variance in well-being and productivity. Results include (1) the quality of social contacts predicted positively, and stress predicted an individual's well-being negatively when controlling for other variables consistently across both waves; (2) boredom and distractions predicted productivity negatively; (3) productivity was less strongly associated with all predictor variables at time two compared to time one, suggesting that software engineers adapted to the lockdown situation over time; and (4) longitudinal analyses did not provide evidence that any predictor variable causal explained variance in well-being and productivity. Overall, we conclude that working from home was per se not a significant challenge for software engineers.
1 Introduction
The study addresses limited longitudinal evidence on factors associated with software engineers’ well-being and productivity during pandemic remote work. It combines broad predictor coverage with a two-wave design to identify variables uniquely associated with these outcomes.
- Pandemic restrictions created substantial well-being and productivity costs, while their effects on software professionals working remotely remained insufficiently understood.
- The study integrates organizational and psychological theories with remote-work research and public health and work recommendations.
- The study included a broad range of variables to identify which predictors were uniquely associated with software professionals’ well-being and productivity.Including multiple relevant variables helps distinguish primary associations from effects potentially driven by related factors.
- A longitudinal design allowed the researchers to examine whether predictors explained well-being and productivity beyond other variables over time.
- The research investigates relevant predictors of well-being and productivity for software engineers working remotely during a pandemic.
2 Related Work
Related work shows that quarantine can harm well-being, while remote work has both benefits and challenges for productivity. Prior evidence leaves uncertainty about productivity during disasters and about whether existing remote-work findings generalize to pandemic conditions.
- Prior well-being studies were mostly cross-sectional, used limited predictors, and left some quarantine-specific mechanisms insufficiently tested.
- Quarantine research documented psychological harms, but evidence on productivity among people who continue working remained limited.Prior work largely focused on mental and physiological health rather than the productivity effects of quarantine.
- Remote work has been associated with better work-life balance and productivity, but also with collaboration difficulties, loneliness, distractions, and difficulty staying motivated.
- Existing remote-work findings may not generalize because remote workers can differ from the broader working population through self-selection.
- Pandemic restrictions intensified ordinary remote-work difficulties by forcing unprepared workers into unfamiliar home-working conditions.
3 Research Design
The research uses a theory-informed, two-wave design covering 51 predictors of software engineers’ well-being and productivity during lockdown. Participants were recruited and surveyed remotely, with the second-wave variables selected partly from first-wave associations.
- Well-being was operationalized as broad subjective life satisfaction, while psychological variables were treated as determinants of overall well-being.
- The study measured 51 predictors across broad psychological, social, situational, and physiological domains in a two-wave longitudinal design.The initial predictor selection was theory- and literature-driven, while the second-wave selection was data-driven.
- The design included coping strategies, volunteering, psychological needs, social-relationship quality, and personality-related factors as potential predictors.
- Power analysis indicated that 190 participants were needed for the planned medium-to-large effect size and .80 power.
- Participants were recruited through Prolific and surveyed with Qualtrics, using screening procedures to identify software professionals working remotely during lockdown.
- The first wave included 192 participants, and 184 completed the second wave, representing 96% participation in wave two.
4 Analysis
The study used a two-stage, two-wave analysis to identify predictors associated with well-being and productivity while accounting for overlapping predictors and measurement reliability.
- Time-1 data identified variables explaining variance in well-being and productivity beyond the other variables.
- The researchers used correlations of at least r = .30 to select predictors for longitudinal testing, with N = 192 and conservative type-I error control.
- The analyses did not transform the data or add unjustified control variables, while demographic checks found no associations with either outcome.
- Multiple regressions tested which correlated predictors explained unique variance in well-being and productivity.
- Structural equation models linked each predictor and outcome across both time points while controlling for outcome autocorrelations and cross-paths.
5.1 Correlations
Well-being showed several substantial correlations, led by stress, social contacts, and autonomy needs, while participants’ interpretations of extraversion during lockdown often conflicted with the observed association.
- At time 1, 16 variables correlated with well-being at r ≥ .30, with the overall pattern consistent with prior literature.
- Stress correlated negatively with well-being at r = −.58, while quality of social contacts and need for autonomy correlated positively at r = .49 and r = .48.
- Generalized anxiety predicted well-being more strongly than COVID-19-related anxiety, with B = −.58 versus B = −.11.
- Extraversion was positively correlated with well-being at both waves, contrary to most participants’ expectations.
- Only 2 participants predicted that introverts would struggle more, whereas 136 predicted extraverts and 46 expected equal difficulty.
- Participant accounts described different social challenges for introverts and extraverts, including difficulty maintaining contact or making one’s presence visible online.
5.2 Unique influence — Multiple regression analyses
Multiple regressions showed that stress and social-contact quality uniquely predicted well-being across both waves, whereas correlated productivity predictors did not separate statistically.
- Multicollinearity was not considered problematic because VIF values remained below 4.1 in all four regression models.
- Stress and quality of social contacts uniquely predicted well-being at both time points, while the full model explained R2 = .44 at time 1 and R2 = .47 at time 2.
- At time 1, stress, social contacts, and daily routines uniquely predicted well-being at α = .05.
- At time 2, competence and autonomy needs, stress, social-contact quality, and sleep quality uniquely predicted well-being at α = .05.
- Four variables jointly explained 16% of productivity variance at time 1 and 8% at time 2, but none explained variance beyond the others.
- Need for competence correlated positively with well-being but became negatively associated in controlled regressions, illustrating sensitivity to included predictors.
5.3 Longitudinal analysis
Longitudinal analyses examined whether predictors at wave 1 explained well-being or productivity at wave 2, using SEM and within-subject comparisons. No predictor significantly explained another variable across time, while well-being and social contacts showed some mean changes over two weeks.
- Structural Equation Modeling: The SEM models involving autonomy, competence, and relatedness sometimes fit poorly and were not discussed further.The authors did not use modification indices to improve model fit in the main analyses.
- Structural Equation Modeling: Stress and well-being were significantly associated within both waves, but stress at time 1 did not significantly predict well-being at time 2.The within-wave coefficients were B = −0.75 at time 1 and B = −0.15 at time 2; the cross-time estimate was B = −0.00, p = .99.
- Structural Equation Modeling: No model revealed significant cross-time associations at α = .005, so wave-1 variables did not significantly explain variance in wave-2 variables.The analysis included 20 SEM models and used a stricter threshold because of multiple testing.
- Mean changes over time: Well-being increased slightly on average from M = 4.14 at time 1 to M = 4.34 at time 2.Ninety-one participants reported higher well-being, 23 reported no change, and 70 reported lower well-being.
- Mean changes over time: Quality of social contacts and behavioral disengagement increased over time, whereas emotional loneliness and communication quality with managers and coworkers decreased.These were reported as average changes across the two waves.
5.4 Exploratory between Gender and Country Analyses
Exploratory comparisons found very few statistically significant differences between groups. Self-distraction differed by gender, while work motivation differed by country.
- Gender analysis: Only self-distraction differed significantly by gender at the .001 threshold, with women reporting higher levels.The comparison covered 65 variables; other differences, including higher anxiety among women, were not statistically significant.
- Country analysis: Only material extrinsic work motivation differed significantly between participants in the United Kingdom and United States at the .001 threshold.Participants based in the United States reported higher average levels.
5.5 Conceptual replication analysis
The replication analysis revisited an apparent discrepancy about office setup by separating distraction-related ergonomics from other ergonomic factors. Distraction-related ergonomics showed stronger associations with well-being and productivity, supporting the importance of distinguishing these constructs.
- Conceptual replication analysis: Ergonomics-distractions correlated more strongly with well-being, r = .25, and productivity, r = .29, than ergonomics-others, with r = .19 for both.The distraction-related measure combined distraction and noise items, while the remaining four items formed ergonomics-others.
- Conceptual replication analysis: The replication suggests that distraction, rather than office setup broadly, may account for the associations previously attributed to ergonomics.The authors state that the findings replicate Ralph et al.’s results while emphasizing the distinction between distraction and office setup.
6 Discussion
Well-being and productivity were associated with multiple factors, while longitudinal analyses did not establish causality; software professionals also appeared to adapt to lockdown over time.
- Findings: Well-being and productivity were positively associated, but the study found no causal relations between the analyzed variables and either outcome.All 20 structural equation models were non-significant, so the direction of influence remains unresolved.
- Findings: Social-contact quality and stress were among the most reliable predictors of well-being, while boredom and distractions were linked to lower productivity.The authors frame these findings as bases for practical recommendations concerning social connection, stress reduction, and the home work environment.
- Implications and recommendations: Structured routines, meaningful social contact, stress reduction, adequate sleep, and designated work areas are presented as practical supports for remote well-being.The recommendations include organizing the day, maintaining relationships, using mindfulness or other stress-reduction activities, and reducing household distractions.
- Implications and recommendations: Introverted professionals may face greater difficulty maintaining contacts remotely because reaching out becomes more proactive and formalized outside the office.The authors connect this difficulty to reduced opportunities for both structured and unstructured interaction with colleagues and friends.
- Findings: Over time, software engineers appeared to adapt to lockdown as well-being increased and perceived social-contact quality improved.The conclusion describes adaptation despite continuing challenges associated with working from home.
- Threats to validity: Interpretation is limited by self-reports, possible omitted variables, unmeasured lockdown-severity perceptions, and insufficient cross-country samples for moderation tests.The study also could not compare its findings with non-remote pre-pandemic settings and could not draw causal conclusions.
7 Conclusion
The study identifies correlates of software professionals’ well-being and productivity during pandemic-related remote work, while finding no evidence that the tested predictors causally explained either outcome. It also emphasizes that intervention effects may vary across individuals and require further validation.
- High stress, absent daily routines, and limited social contacts were among the variables most related to well-being, while boredom and distractions were related to lower productivity.
- A longitudinal study of 192 software professionals combined correlations, multiple regressions, and structural equation modeling across two waves.The study tested 51 literature-derived variables and found several significant unique predictors, but no significant causal relations in 20 structural equation models.
- Well-being increased on average during the pandemic, and well-being was correlated with productivity; nine of 51 factors were reliably associated with the two outcomes.
- The findings motivate actionable recommendations, but organizations should test them incrementally because intervention effects are unlikely to be identical across people.The paper recommends gathering employee feedback and considering individual differences when evaluating interventions.
- Future studies should test generalizability with representative cross-sectional samples and examine mechanisms and software-tool effects on well-being and productivity.
A Appendix
The appendix documents the study’s instruments, reliability information, statistical models, longitudinal comparisons, demographic comparisons, and predictor correlations. It also includes figures and tables supporting analyses of well-being and productivity across two waves.
- Figures 3–8 visualize within-subject comparisons and regression coefficients for well-being and productivity at times 1 and 2.
- Table 9 summarizes constructs, instruments, reliability estimates, and instrument changes, including measures of loneliness, anxiety, stress, boredom, routines, motivation, social contacts, and distractions.
- The appendix reports instruments for personality, psychological needs, work motivation, physical activity, sleep, social relationships, communication, and home distractions, with reliability notes and adaptations.
- Structural equation modeling results are organized by independent and dependent variables and report regression estimates, standard errors, p-values, and fit indices including CFI, RMSEA, and SRMR.
- Additional tables report within-subject changes over time, comparisons by gender and country, and correlations of well-being and productivity with predictors at times 1 and 2.