What does a survey of 27,792 hourly workers say about unstable schedules and well-being?

Across the scheduling practices Schneider and Harknett (2019) measured, canceled shifts separate the two groups of workers by more than any other. Among the 14% of workers who had a shift canceled, 64% reported psychological distress, versus under half of those who did not. On sleep, 82% reported poor sleep quality, versus 72%. On the happiness item, 43% reported being not too happy, versus 26%.

Two features of those figures matter more than their size. They are descriptive percentages comparing two groups of survey respondents, not regression-adjusted effects, so nothing has been held constant between the workers who lost a shift and the workers who did not. The happiness figure also comes from the bottom band of a standard three-point happiness item, which is why the accurate phrasing is that 43% reported being not too happy, versus 26%, rather than any looser statement about how unhappy the two groups were.

There is a plausible operational reading of why cancellation sits at the top, though the survey itself cannot establish it. A canceled shift withdraws income the worker has already planned around, after childcare, transportation, and often a second job have been arranged to fit it. It also reverses a commitment the employer already made, which is different in kind from a schedule that goes up late. That interpretation is worth holding lightly, because it is reasoning about the finding rather than part of the finding.

The paper also reports simulations. The authors model that eliminating on-call shifts would be associated with about 15 percentage points less psychological distress, 8 points better sleep quality, and 9 points more happiness among affected workers, and that a 72-hour advance-notice requirement would be associated with nearly 5 points less distress. The on-call simulation applies to the 26% of the sample who reported working on-call shifts.

These are modeled projections built on cross-sectional data, and the authors state plainly that the estimates are not causal. The simulations describe how the surveyed population would look on the measured outcomes if the on-call group resembled everyone else. They are not a forecast of what happens at your sites if you retire on-call scheduling next quarter. Read them as the size of a gap worth investigating, not as a return you can put in a plan.

The sample is large and pointed squarely at the population most exposed to unstable scheduling: 27,792 hourly workers at 80 of the largest US food-service and retail firms. It is also a non-probability sample, recruited through targeted Facebook and Instagram advertising, so the percentages describe the workers who responded rather than the national hourly workforce. That is a limit on generalization, not a reason to dismiss the pattern.

Data collection ran from June 2016 to October 2017. Service-sector scheduling conditions shifted after 2020, through tighter labor markets, fair-workweek ordinances in more jurisdictions, and broader adoption of scheduling software, so the levels reported here belong to that window and should be quoted with the dates attached.

The direction of the relationship is the sharper problem for this particular finding, because distress plausibly shapes who loses a shift rather than the other way around. A worker managing anxiety or a sleep deficit may call out more often, ask to be cut early, or read as less dependable to the manager deciding whose shift to pull on a slow afternoon. On that account the distress comes first and the canceled shift follows it. A survey that captures the schedule and the distress in the same sitting cannot tell that story apart from its opposite, and the paper does not claim otherwise. Use the study for what it does well, which is showing where the largest well-being gaps sit among practices employers directly control, and look to experimental evidence when you need to know what changing a practice produces.