Estimates of how much money moves through online labour platforms depend on what is counted as earnings, whose records are used, and how far respondents recognise the activity when a survey asks about it. Three strands of published work — bank-transaction studies, survey redesigns, and tax-administrative analysis — have produced numbers that do not line up neatly. That misalignment is not treated here as a puzzle to be resolved; it is the subject of the research summarised below.
Bank records and platform deposits
Diana Farrell and Fiona Greig, writing for the JPMorgan Chase Institute, analysed de-identified checking-account deposits identified as coming from a set of online platforms. Their reports on the Online Platform Economy described participation rates and deposit volumes among the bank’s customers over multi-year windows. The strength of the approach is frequency: deposits can be observed month by month without relying on recall. The limitation, which the authors state, is selection: the sample is customers of one bank, platform identifiers must be mapped from payee strings, and people who route platform income elsewhere are not observed. The series therefore speaks to patterns inside that customer base rather than to a census of platform work in the United States.
Household surveys and recognition
Katharine Abraham and coauthors have examined how household surveys capture online platform work and related forms of nontraditional employment. Their work shows that standard employment questions can miss intermittent or secondary platform activity when respondents do not classify it as a job in the sense the questionnaire expects. Experimental and redesigned items improve detection in some settings, but the improvement depends on wording, reference period and whether the activity is treated as self-employment. Abraham and colleagues emphasise that undercount is not uniform: it varies by platform type and by how central the activity is to household income.
Tax files and Form 1099-K
Brett Collins, Andrew Garin, Emilie Jackson, Dmitri Koustas and Mark Payne have used United States tax records, including information returns associated with online platforms, to study participation and earnings. Administrative files can reveal activity that surveys miss, particularly when amounts are small or episodic. The authors also document limits: reporting thresholds, changes in Form 1099-K rules, and the difference between gross receipts and net earnings after expenses. Tax data measure what was reported to tax authorities under the rules then in force; they do not, by themselves, measure take-home pay after costs that never appear on the information return.
What the strands do not settle
Taken together, the bank, survey and tax literatures agree that platform-related income exists at a measurable scale for some populations and that measurement choices change the estimated size. They do not agree on a single headline rate that can be transplanted from one data source to another without restating the population, the period and the definition of earnings. Farrell and Greig’s deposit series, Abraham’s survey findings and Collins and colleagues’ tax-based estimates answer different questions. Summaries that blend them into one figure without those qualifications overstate what any of the studies claims.
This note does not recommend a preferred measure. It records that published research treats measurement itself as contested, and that the contest is methodological rather than rhetorical.
Sources named in the text · September 2026