Between 2012 and 2018 researchers at the JPMorgan Chase Institute examined deposit records from a large sample of Chase checking accounts, searching for payments that originated from 128 identified online platforms. The resulting series, published as The Online Platform Economy in 2018 by Diana Farrell, Fiona Greig and Amar Hamoudi, is one of the few descriptions of platform participation that does not depend on people remembering and reporting their own activity in a survey.
The headline measurements are smaller than the surrounding public conversation had suggested. In March 2018, 1.6 per cent of accounts in the sample received a payment from a platform during that month. Across the preceding twelve months, 4.5 per cent received at least one such payment. The distance between those two figures is itself the finding: participation appeared in the records far more often as an episode than as a standing arrangement.
Duration, measured rather than recalled
The duration data points the same way. In transportation, the largest of the four sectors the authors tracked and roughly 63 per cent of all participants, 58 per cent of accounts recorded platform payments in three months or fewer across a full year. In the remaining sectors, fewer than one account in five recorded payments in more than half the months.
Share of total deposits behaves the same way once the frame widens. In a month when an account did record platform payments, those payments represented about 54 per cent of take-home deposits. Averaged across the whole year for every account that participated at all, the same measure fell to roughly 20 per cent. One number describes an active month; the other describes a year containing a few of them. Reporting either without the other produces a different impression of the same dataset.
The series also recorded direction of travel. Transportation-sector payments per active account fell 53 per cent from their 2014 peak by early 2018, while the leasing sector grew 69 per cent over a comparable window and the selling and non-transport service sectors stayed broadly flat. The aggregate category was not moving as one thing.
What the authors said the data cannot show
The report's own limitations section is unusually direct, and any summary that omits it misrepresents the work. Five constraints are named.
The records capture deposits into Chase accounts only. Payments routed to prepaid cards, to accounts at other institutions, or through rails the institute did not observe are absent from the count entirely. The platform list, at 128 entries, may not include every marketplace meeting the authors' criteria, and platforms that were small or new during the window are the likeliest omissions.
The sample skews. It over-represents younger account holders, male-headed families and households in the western United States relative to the national population, which means the participation rates are not national estimates. Roughly 80 per cent of accounts were not continuously present in the sample across the whole study period, so the series is not a fixed panel followed from beginning to end.
The last constraint concerns the unit of observation. Accounts are shared. The analysis can establish that a family received a platform payment; it cannot establish which member of that family performed the underlying activity.
The shape of the claim
Administrative data of this kind answers a narrow question well. It records that money arrived, from whom, and when. It does not record hours worked, effort, motivation, what happened outside the observed accounts, or why participation started or stopped. Farrell, Greig and Hamoudi frame their conclusions accordingly, and the frame is part of the result.