Labor Economics
Output Per Hour in the Platform Economy: What Three Studies Found
Measuring how much a person produces per hour of work is a cornerstone of labor economics. For traditional employment, the Bureau of Labor Statistics publishes quarterly labor productivity data based on output and hours worked. For the platform economy, no equivalent standard exists, and the studies that have attempted to fill this gap have used methods that are not directly comparable.
Study one: Hara and colleagues on effective wages
Kotaro Hara, Abigail Adams, Kristy Milland, Saiph Savage, Chris Callison-Burch, and Jeffrey Bigham published a study at the 2018 CHI Conference on Human Factors in Computing Systems that attempted to measure the effective hourly wages of Mechanical Turk workers. By analyzing over 2,600 workers completing over 3.8 million tasks, the research team calculated earnings per hour by dividing total earnings by total time spent on the platform, including time spent searching for and evaluating tasks.
The study found that the median effective hourly wage was substantially below the U.S. federal minimum wage. Critically, the researchers noted that the gap between posted task rates and effective hourly earnings was largely explained by unpaid time — the time workers spent on the platform without actively completing paid tasks. This finding is consistent with the non-task labor documented in ethnographic research on the same platform.
A methodological consideration noted by the authors is that their time-tracking approach relied on browser-based monitoring tools, which may not capture all idle time or may misclassify certain browsing activities as task-search time. The direction and magnitude of this potential bias were not fully resolved in the study.
Study two: the Online Labour Index approach
Vili Lehdonvirta and colleagues at the Oxford Internet Institute took a different approach with the Online Labour Index (OLI), first published in 2017. Rather than measuring individual worker throughput, the OLI tracks the volume and type of digital work available across major platforms, using automated data collection to monitor task postings in real time.
The OLI provides a macro-level view of how much work is being posted and in which categories — software development, creative, clerical, writing, and others — but it does not directly measure how quickly tasks are completed or what individual workers earn per hour. Lehdonvirta has noted in subsequent publications that the OLI is designed as a market-level indicator, analogous to economic indexes that track job postings in the traditional labor market. It captures demand for digital labor, not the supply-side experience of completing it.
This distinction is important because throughput at the market level and throughput at the individual level can move in opposite directions. If a platform adds more workers while the volume of available tasks remains constant, individual throughput may decline even as aggregate task completion remains stable or increases.
Study three: Katz and Krueger on alternative work
Lawrence Katz of Harvard University and Alan Krueger of Princeton University published a widely discussed study in the American Economic Review: Papers and Proceedings in 2019, examining the prevalence of alternative work arrangements in the United States. Their data, drawn from the RAND American Life Panel survey, included questions about online platform work.
Katz and Krueger estimated that the share of workers engaged in alternative work arrangements — including independent contractors, on-call workers, and temporary help agency workers — had increased from 10.7 percent in 2005 to 15.8 percent in 2015. However, they also reported that subsequent analysis using the Bureau of Labor Statistics Contingent Worker Supplement in 2017 did not confirm the magnitude of this increase, raising questions about whether the original survey instrument or sampling method inflated the estimate.
The relevance to output-per-hour measurement is indirect but significant. If the size of the platform-work population is itself uncertain, then any estimate of aggregate output or average productivity derived from that population estimate inherits the same uncertainty. Katz and Krueger acknowledged the discrepancy between their survey results and the BLS findings and suggested that differences in survey design, question wording, and reference periods may account for part of the gap.
What the divergence means
The three studies described here approach the question of output in the platform economy from different levels of analysis — individual, market, and population — and with different instruments. Hara and colleagues measured what workers actually earn per hour of platform time. Lehdonvirta and the OLI team measured how much work the market generates. Katz and Krueger measured how many people participate in non-traditional work arrangements.
None of these alone provides a complete picture of output per hour in the platform economy. Each captures a real dimension of the phenomenon, but combining their findings requires careful attention to the different definitions, populations, and time frames involved. For the moment, the academic literature offers multiple partial measurements rather than a single authoritative figure, and published numbers should be interpreted within the context of the methodology that produced them.