Charted: Platform Earnings Swing Month to Month — For Whom Is Gig the Majority Paycheck?

Gig-platform income is sold as flexible side money. Bank-transaction ledgers tell a sharper story: the swing is the product. JPMorgan Chase Institute (JPMCI) administrative data show labor-platform participants are active in only 56% of months after they first earn — so nearly half of months print $0 platform income after a month that averaged $533. In those active months, platform earnings were about 33% of total income. Capital platforms are thinner still: active 32% of months, $314 average, 20% of income when working.

The household question is not whether platforms exist. It is how often the paycheck disappears, and for what share of participants platforms are the majority of earnings. As of September 2015, 25% of active labor-platform participants relied on platforms for more than 75% of total income — a majority-paycheck tail — while 46% cleared a 25% reliance bar. Pandemic-era updates put the median transport driver near 10% of family income from platforms: secondary for the middle of the distribution, essential for a thick left tail. Fed SHED 2024 adds the survey layer: only 21% of adults who did gig activities called them their main job, yet 61% of platform-task participants wished pay were more consistent.

The dashboard above is built for that tension. Toggle Earnings swings, Reliance bands, Participation path, Sector medians, Substitute vs supplement, and Fed SHED attitudes. Filter labor vs capital on the reliance panel. The rest of this post is the narrative behind those panels.

The swing is mostly on/off, not a smooth W-2 wobble

Start with the ambient cash-flow weather. In JPMCI’s million-customer sample (October 2012–September 2015), individuals saw an average 40% absolute month-to-month change in total income; 55% experienced changes greater than 30%. Labor income still dominated the volatility budget because it is most of the paycheck. That is the floor under every platform comparison: Americans already live with large MoM swings inside traditional jobs.

Platform earnings layer a binary switch on top of that floor. After first participating, labor-platform earners were active only 56% of months; capital-platform earners only 32%. The implied MoM swing from an active labor month ($533) to an idle month ($0) is a 100% drop in platform income — not a 5% overtime wobble. Capital platforms show the same on/off pattern at a lower average ($314). That intermittency is why monthly participation counts understate annual exposure: JPMCI’s pandemic-era path peaked near 2.5% of families earning platform income in a given month, but about 7.5% had earned some platform income in the prior year — a five-point gap between the “this month” census and the “this year” census.

Within active months, demand shocks still bite. Among drivers who kept participating, median platform revenues fell about 40% in April 2020 before rebounding — a within-active MoM crash layered on top of a one-third drop in driver participation. Selling platforms kept growing; leasing and transport took the hit. The sector panel is the operational map of which platform “jobs” swung hardest.

Reliance: secondary for the median, majority for a thick tail

Headline averages can hide the household that treats Uber or DoorDash as rent money. JPMCI’s September 2015 cut of active participants is the cleanest public reliance ladder:

Among everyone who had ever participated over three years — including months they were idle — 82% of labor participants and 96% of capital participants relied on platforms for less than 25% of income in the September 2015 snapshot. Growth in headcount did not deepen reliance: activity rates and income shares stayed roughly flat even as cumulative participation exploded. The platform economy scaled by recruiting new entrants, not by converting side hustles into primary jobs for the typical participant.

The pandemic-era family cut reframes the same idea with sector medians. The median transport driver’s platform share of family income settled near 10% after the onset of COVID; non-transport work drifted toward 7%; sellers stayed around 1–2%; lessors sat highest, roughly 15–20%. Median is not mean, and median is not the >75% tail. Briefings that say “gig is just a side hustle” are describing the middle of the distribution. Briefings that say “gig is the new job” are describing the left tail — and JPMCI shows that tail is large enough to matter for credit, UI eligibility, and cash-flow stress.

Labor substitutes; capital supplements

Platforms are not one labor market. In months with labor-platform earnings, non-platform income was about 14% lower and labor-platform earnings contributed about 15% of income — so total income barely moved. Labor platforms were filling a hole. Capital platforms (selling goods, leasing assets) behaved differently: non-platform income was almost unchanged, and platform earnings added about 7%, lifting total income by roughly 7%. That is the substitute-versus-supplement panel: rides and tasks as gap fillers; Airbnb and marketplace sales as top-ups.

The Fed SHED 2024 survey rhymes without using bank wires. 20% of adults did some gig activity in the prior month, but only 4% did short-term tasks arranged through an app or website. Among people who did gig activities, just 21% called them their main job, and 96% usually spent under 35 hours a week on them. Yet 31% said that without gig income they would have trouble making ends meet — 41% among platform-task adults — and 61% of platform-task participants wished the pay were more consistent. Flexibility scores high (78% for platform tasks); consistency does not. That is the consumer-finance signature of volatile platform earnings: optional on the calendar, non-optional when the rent is due.

What BLS can and cannot say here

The BLS Contingent Worker Supplement measures a different object. In July 2023, 4.3% of workers held contingent jobs as their sole or main job — temporary or expected to end. The May 2017 electronically mediated employment questions captured app-arranged short tasks in a single reference week and remain a useful perimeter check, not a MoM earnings ledger. Survey main-job rates will understate bank-observed intermittent platform deposits; bank samples will over-represent Chase customers. This post keeps those perimeters labeled: JPMCI for high-frequency earnings and reliance, SHED for attitudes and broad gig prevalence, BLS for contingent / electronically mediated stock.

Caveats and confidence

Activity rates, average earnings, and >75% / >25% reliance shares for labor and capital platforms are disclosed JPMCI findings from the 2012–2015 Chase sample and should be treated as that vintage — the cleanest public MoM ledger, not a 2026 census. Pandemic participation peaks (2.5% monthly / ~7.5% trailing year), the −25% MoM participation drop, driver revenue and income-share medians follow JPMCI’s 2018–2021 family sample; chart paths in the dashboard include desk interpolations between disclosed anchors and are tagged estimated where needed. SHED 2024 gig definitions are broader than online labor platforms (selling used clothes counts); platform-task results use the half-sample aligned to Contingent Worker Supplement wording. Chase administrative data are not a probability sample of all U.S. adults. None of those caveats erase the dual pattern: platform income is intermittent by design, and majority reliance is a minority — but a large one — of active labor participants.

Reading the dashboard for a briefing

Start on Earnings swings to see why idle months dominate MoM platform volatility. Flip to Reliance bands and filter Labor to put the 25% majority-paycheck tail on screen. Use Participation path for the monthly-versus-trailing-year gap and the COVID participation crash. Sector medians separates drivers from sellers and lessors on both dollars and income share. Substitute vs supplement is the one-slide labor-versus-capital policy distinction. End on Fed SHED attitudes when the room needs the “wish pay were consistent” and ends-meet numbers. The source note under the charts carries the full methodology string for citation.