The American Time Use Survey (ATUS), conducted annually by the Bureau of Labor Statistics since 2003, asks a nationally representative sample of Americans to recount every activity from a single 24-hour day. Respondents describe what they did, when, for how long, and where. When researchers use these diaries to study home-based goods production — including craft work — the estimates they produce depend heavily on which activity codes they include and how they define the boundary between production and leisure.
The coding problem
ATUS activities are classified using a hierarchical coding system. "Arts and crafts as a hobby" falls under the leisure category. "Working at a job held at home" falls under the work category. A person who spends two hours sewing handbags for sale on a marketplace platform might code the activity as work, as a hobby, or as something else entirely, depending on how they frame the activity to the interviewer. The distinction between income-generating craft production and personal craft leisure is not always clear in the respondent's own mind, let alone in the data.
Economists Mark Aguiar and Erik Hurst, in their widely cited 2007 paper on leisure trends in the United States, included craft and hobby time in their broad definition of leisure. Their framework was not designed to separate commercial craft production from recreational craft activity — a distinction that matters considerably when the goal is to measure how much time people spend on income-generating handmade work. Researchers who reuse the Aguiar-Hurst categorization inherit this limitation.
Boundary activities
The challenge intensifies for activities that sit between production and administration. Photographing finished items, writing product descriptions, packaging orders, and responding to customer messages are all part of running a craft business, but they are coded differently in ATUS depending on how the respondent describes them. Benjamin Bridgman, working at the Bureau of Economic Analysis, examined the broader category of home production in a 2016 paper and noted that the line between market work done at home and non-market household production is one of the most persistent measurement problems in time-use research.
For craft producers specifically, the ambiguity is compounded by the fact that many of the same physical activities — cutting fabric, mixing glazes, assembling components — can be either market production or personal consumption depending on the maker's intent. ATUS does not ask about intent directly; it relies on the respondent's description and the interviewer's coding judgment.
The single-day limitation
Because ATUS captures only one day per respondent, it provides a snapshot, not a pattern. Craft work is often episodic: a producer might spend an entire Saturday building inventory and then not touch the workshop for two weeks. A single diary day has roughly a one-in-fourteen chance of landing on a production day for someone who works on crafts every other weekend. Valerie Ramey and Neville Francis, in their 2009 analysis of long-run trends in time use, noted that single-day diaries can understate activities that are concentrated on particular days of the week, though their focus was on broader labor categories rather than craft production specifically.
What this means for estimates
The combined effect of coding ambiguity, boundary-activity classification, and single-day sampling means that ATUS-based estimates of craft production time should be interpreted as rough indicators rather than precise measurements. A study reporting that Americans spend an average of a certain number of minutes per day on home-based goods production is reporting a population-level average that smooths over the highly skewed distribution of actual craft activity — where most respondents report zero minutes and a small number report several hours.
None of this makes ATUS data useless, but it does mean that claims about "how much time" people spend on craft-based production are only as good as the coding decisions and boundary definitions that produced them. Researchers who acknowledge these constraints — as Bridgman and others have done — provide more useful context than those who present time-use averages as settled facts.