Part 03
Electronically mediated employment in US contingency measures
In 2017, the US Bureau of Labor Statistics fielded questions in the Contingent Worker Supplement (CWS) to the Current Population Survey that asked about electronically mediated employment — work obtained through a company website or mobile app that connected the worker directly with customers and arranged payment. The BLS later published estimates describing how many people reported such work and how it overlapped with contingent and alternative employment arrangements.
The instrument mattered because it tried to capture platform-style matching inside a household survey rather than inside a single company’s administrative file. Respondents were asked about work in a recent period, with wording intended to distinguish app- or website-mediated customer matching from ordinary online job boards that simply advertise vacancies for traditional employers.
What the estimates showed — carefully
Published BLS materials described electronically mediated work as relatively uncommon among the employed in the measured period, with further breakdowns by whether the work was a main job or additional work. Exact headline percentages belong to the official tables and should be cited from those tables when used in research; this note focuses on the measurement framing rather than on repeating a single percentage out of context.
Overlap with contingent work concepts is important. Someone can perform electronically mediated work as an independent contractor, as short-term payroll labor, or in arrangements that do not map cleanly onto older contingency categories. The CWS questions were designed to sit alongside those older categories, not to replace them.
Instrument limits
Household respondents may under-report short, irregular platform tasks, especially when those tasks are secondary to a payroll job and were not active in the reference window. Respondents may also over-include work that used a website only for finding clients without intermediary matching and payment. Cognitive testing and subsequent methodological commentary have discussed how wording choices affect classification.
Another limit is vintage. Digital platforms change product design faster than survey supplements are re-fielded. An estimate from a single supplement year is a snapshot of how respondents understood the questions at that time, not a continuous index of “gig economy size.” Later research often triangulates CWS-style measures with tax forms, private datasets or newer survey modules.
How this part relates to the others
Hara and colleagues measured earnings distributions on one microtask platform with scraped activity logs. Multiple-jobholding rates count second jobs in a reference week without requiring platform identity. Electronically mediated employment questions ask whether matching and payment ran through a website or app. Each answer is valid for its object and incomplete for the others.
Cue Ledger keeps the three notes adjacent so readers can see that “side income research” is not one number. It is a family of instruments that sometimes share vocabulary while measuring different economic acts. Confusing those instruments produces claims that no single source supports.