Research on working hours that change from week to week

Work that pays by the hour often comes with hours set at short notice. The research describing this ranges from careful documentation to one randomised experiment inside a retail chain.

Documenting the practice

Susan Lambert and Julia Henly, working at the University of Chicago, spent years documenting how hours are assigned in hourly work: how far ahead a schedule is posted, how far the number of hours moves between one week and the next, and how often a shift is held open on call. With Peter Fugiel, they analysed early-career workers in the National Longitudinal Survey of Youth and reported that a substantial share of hourly employees learned their schedules roughly a week ahead or less. Their work describes an employer practice and the reasons employers give for it, rather than the circumstances of any individual worker.

A recurring point in that research is that scheduling variation is not a single thing. Advance notice, week-to-week variation in the number of hours, cancelled shifts, on-call requirements and split shifts are separable practices that tend to be measured together and that need not appear in the same workplace. Studies that combine them into one index produce results that are difficult to compare with studies that do not, which is one reason findings across this literature are harder to line up than the headline summaries suggest.

Surveys of service workers

Daniel Schneider and Kristen Harknett built the Shift Project to survey workers in retail and food service directly, recruiting respondents through targeted advertising on social platforms. Their published papers report associations between unstable scheduling and measures of sleep quality and psychological distress. The authors are explicit about what this design can and cannot support. Recruitment through advertising produces a non-probability sample that has to be weighted against known benchmarks. The measures are self-reported. Most of the analyses are cross-sectional, which means an association is observed without the direction of influence being established.

An experiment rather than an association

Joan Williams, Saravanan Kesavan and Susan Lambert ran a rarer kind of study: a randomised intervention inside stores of a single retail chain, in which some locations adopted more stable scheduling practices and others continued as before. The published results reported gains in sales and in labour productivity at the treated locations. The constraints are equally clear. The study covered one retailer over a limited window, the number of participating stores was small for a field experiment, and several components of the intervention were adopted unevenly across locations, which makes it difficult to attribute the result to any single change.

Local ordinances and their evaluations

Several United States cities, including San Francisco, Seattle, New York and jurisdictions in Oregon, have passed scheduling ordinances that require advance notice of hours and compensation when hours change late. Harknett, Schneider and co-authors have published evaluations comparing outcomes before and after these rules took effect, reporting improvements in how far ahead schedules were posted in Seattle. Evaluations of this kind carry the usual difficulty of separating the rule from everything else changing in a local labour market at the same time, and results from one city are not straightforwardly transferable to another.

What remains unsettled

The largest open question in this literature is selection. Workers are not assigned at random to jobs with unstable hours, so an association between unstable hours and a health measure may partly reflect which people hold those jobs in the first place. The field experiment addresses that problem for one employer and one period. Whether its findings extend to other sectors, other countries or other economic conditions has not been established, and the researchers involved have generally said so themselves.

Back to the library