What is our question?
For pre-retirement New Zealand adults, how would shifting weekly work hours by ten hours affect 28 well-being outcomes one year later? We compare each shift, upward and downward, with the natural course.
Target trial emulation · NZAVS 2020–2023 · Lab report
Work Hours, Fatigue, and Physical Health
How work hours affect well-being: a target trial emulation in a national New Zealand cohort
Work hours are associated with differences in well-being. We ask how well-being would change if weekly work hours increased or decreased. We emulate a target trial using data from 24,579 pre-retirement participants in the New Zealand Attitudes and Values Study (NZAVS). We estimate how 28 well-being outcomes would differ under interventions that shift weekly work hours ten hours up or ten hours down, each compared with the natural course: work hours changing as they ordinarily would.
24,579pre-retirement adults, NZAVS
3annual waves, 2020/2021–2022/2023
28well-being outcomes
±10hours per week, the two policies
The emulated trial
A target trial emulation organises an observational study around the question and design of a hypothetical randomised trial.
What is our population?
The work-hours analysis includes 24,579 NZAVS participants of pre-retirement age at baseline (2020/2021). Work hours are measured in 2021/2022 and outcomes in 2022/2023. Censoring adjustment accounts for people leaving the study, aiming to estimate outcomes for the baseline cohort under complete follow-up.
What interventions do we compare?
Shift up · +10 hours
Increase weekly work hours by ten, compared with the natural course, which allows work hours to change as they ordinarily would.
Shift down · −10 hours
Decrease weekly work hours by ten, compared with the natural course, which allows work hours to change as they ordinarily would.
Can the data support the comparison?
Some changes in hours are easier to study
To estimate the effects of changing work hours, we need evidence about comparable people working at the hours the intervention would produce. Positivity requires that any set of work-hour values given a positive chance by an intervention also has a positive chance among people with each relevant history. Here, history means the characteristics and earlier survey responses used to compare participants. In practice, we also need enough comparable observations to reliably estimate the effects.
The observed work-hour distributions provide better coverage for the ten-hour reduction. Increasing hours by ten more often produces work-hour values that are rare among people with comparable measured histories. Estimating the upward shift therefore requires more extrapolation: using models to predict outcomes where comparable observations are sparse. That reliance calls for greater caution.
The central result
Increasing work hours raises estimated fatigue; decreasing hours lowers it.
Estimated effects of the ten-hour shifts concentrate in tiredness, measured as fatigue, and a small set of physical-health outcomes. Most point estimates, our best estimates of the average changes, are small under either policy. Moreover, the data provide stronger empirical support for decreasing hours than for increasing hours. The upward shift therefore warrants greater caution. Comparisons between people at the first survey suggest differences across more outcomes and sometimes point in the opposite direction. However, those associations describe differences between people. The adjusted longitudinal estimates use later outcomes to address the specified changes in work hours.
+10 hours · estimated changes
Higher fatigue and less sleep
As noted above, increasing weekly work hours by ten raises estimated fatigue by 0.08 standard deviations, a small change relative to the variation in fatigue across people. Estimated nightly sleep falls by roughly 3.4 minutes. Body mass index increases by roughly 0.19 kg/m² and perceived physical health declines by roughly 0.13 points. However, estimates for sleep, body mass index, and perceived physical health warrant greater caution about uncertainty or residual confounding, distortion from common causes of work hours and well-being that remains after adjustment. The estimated increase in perceived support is also small and sensitive to residual confounding.
−10 hours · estimated changes
Lower fatigue
As noted above, decreasing weekly work hours by ten lowers estimated fatigue by 0.04 standard deviations. Estimated body mass index falls by 0.13 kg/m², while perceived physical health increases by 0.07 points. However, both physical-health estimates remain sensitive to residual confounding.
The estimates
Estimated average effects of increasing or decreasing weekly work hours by ten, each compared with the natural course. Effects are in standard-deviation units, which express each change relative to the variation in that outcome across people. Bonferroni-adjusted intervals control family-wise error at 5% across 28 outcomes under each policy (approximately 99.82% individual intervals). This adjustment accounts for testing several outcomes together. Green marks the fatigue estimate under each policy, which the paper identifies as least vulnerable to unmeasured confounding. Gold marks estimates whose corrected intervals exclude zero but whose E-value bounds are modest. E denotes the E-value for the confidence limit closest to zero. Larger E-values mean that stronger unmeasured confounding would be needed to move that limit to zero. The paper's supplement reports all 28 outcomes.
Inference & robustness
01
The natural course comparator
As stated above, we compare each ten-hour shift in weekly work hours with the natural course, allowing work hours to change as they ordinarily would. Censoring adjustment targets both comparisons to the baseline cohort under complete follow-up.
02
Adjustment
We estimate the work-hour intervention effects with modified treatment policy estimators in the lmtp R package. Machine-learning models adjust for baseline differences and dropout. Adjustment addresses confounding: common causes of work hours and well-being that can distort their apparent relationship. Residual confounding is the distortion that remains after adjustment. Causal interpretation depends on the study's assumptions about confounding, exposure support, comparable treatment versions, and missing data.
03
Uncertainty
Uncertainty about the work-hour effects remains substantial. Many adjusted point estimates are closer to zero than the baseline associations. Moreover, many Bonferroni-adjusted intervals include zero, leaving uncertainty about the direction of those effects.
04
Scope
The work-hours estimates concern ten-hour shifts among pre-retirement adults. People gain and lose hours for different reasons, which may affect how the results apply to a particular workplace change. Moreover, work hours and outcomes are self-reported. Reporting errors may bias the estimates in a direction that depends on their relations with exposure, outcomes, and confounders.
What this means
Changing work hours appears to affect how tired people feel. A ten-hour increase in the working week raises estimated fatigue, while a ten-hour reduction lowers it. Most estimated changes in other aspects of well-being are small. The data provide better coverage of the work hours needed to estimate a reduction. The results describe average effects among pre-retirement adults and depend on the study's assumptions. For discussions about working time, fatigue is the clearest outcome to consider; broader benefits remain uncertain.
Citation
Chong, B. X. S., Sibley, C. G., & Bulbulia, J. A. (2026). How work hours affect well-being: A target trial emulation. PLOS One. doi:10.1371/journal.pone.0350816
Data: the New Zealand Attitudes and Values Study. Estimation: modified treatment policies with the lmtp R package.