Survey Methods
What Time-Use Data Reveals About the Structure of Micro-Task Work
Time-use surveys are among the most detailed instruments available for understanding how people allocate their hours across work, leisure, and other activities. When applied to digital micro-task labor, however, these instruments encounter structural limitations that researchers have been working to understand and address.
How the ATUS records work time
The American Time Use Survey (ATUS), administered by the Bureau of Labor Statistics since 2003, asks respondents to report their activities for a single 24-hour day in minute-by-minute detail. The survey uses activity codes to classify each reported segment, with categories for paid work, job searching, education, and dozens of other activities.
Daniel Hamermesh, an economist at Barnard College and the IZA Institute of Labor Economics, has written extensively on the economics of time allocation. In his work on time-use data, Hamermesh has noted that the granularity of the ATUS depends on how respondents describe their activities to interviewers. If a respondent reports that they spent an hour on the computer doing various short tasks for pay, the interviewer must decide how to code that time — and the available activity codes were designed with traditional employment structures in mind.
Where the codes fall short
The challenge is that micro-task work does not fit neatly into the ATUS coding framework. A person who completes twenty data-labeling tasks over the course of an afternoon, interspersed with checking email and scrolling social media, presents a coding problem: is the entire block work time, or only the minutes actively spent on tasks? The ATUS relies on the respondent's subjective account, and there is no mechanism to verify the allocation against platform records.
Katharine Abraham, an economist at the University of Maryland and former Commissioner of Labor Statistics, has examined this measurement gap directly. In a 2018 NBER working paper co-authored with John Haltiwanger, Kristin Sandusky, and James Spletzer, Abraham and colleagues explored the discrepancy between survey-based and administrative measures of employment, finding that informal and non-traditional work arrangements are systematically underreported in household surveys. The research suggested that micro-task and gig work may be particularly affected, because respondents who earn small amounts through these activities may not consider them worth reporting as work.
The probing problem
In a separate line of research, Abraham and Ashley Amaya published a study in the Journal of Official Statistics in 2019 examining whether adding follow-up questions — probes — to the Current Population Survey could capture informal work that standard questions miss. Their experiment found that adding a question about informal, freelance, or online work for pay did identify additional workers who had not reported this activity in response to standard employment questions.
However, the probing approach has its own measurement implications. Abraham and Amaya noted that the additional questions may change how respondents think about their time, potentially leading them to reclassify activities they would otherwise not have considered work. This introduces the possibility that probes capture genuine underreporting in some cases and create overreporting in others. The study did not resolve this ambiguity but recommended further testing to establish the net effect.
Cross-national comparisons
The measurement challenges documented in U.S. data also appear in time-use research conducted in other countries. The Multinational Time Use Study (MTUS), maintained by the Centre for Time Use Research at University College London, harmonizes time-diary data from dozens of countries. Researchers using the MTUS for cross-national comparison of digital work patterns face an additional layer of complexity: the activity codes and interview protocols differ across national surveys, so apparent differences in time spent on digital tasks may reflect coding differences rather than behavioral ones.
Jonathan Gershuny and Oriel Sullivan, who directed the Centre for Time Use Research, have published work on the harmonization challenges involved in making time-use data comparable across countries and time periods. Their research underscores that even within a single country, changes to survey methodology can produce artificial trend breaks that may be mistaken for real changes in behavior.
Implications for throughput measurement
For researchers interested in how much time people actually spend on digital micro-tasks, time-use surveys offer a population-level perspective that platform data alone cannot provide. Platform records capture what happens while a worker is logged in, but they do not capture the full context — the time spent deciding whether to work, the interruptions, the multitasking. Time-use diaries, in principle, capture all of this. In practice, however, the evidence reviewed here suggests that the capture is incomplete and subject to systematic biases that have not been fully quantified. Any estimate of micro-task time use drawn from survey data should be read with these measurement conditions in mind.