Note 02
Uber driver-partner earnings and time at work
Hall and Krueger describe how Uber's driver-partners earned and how they used their time. Gross hourly earnings and utilization are reported together; costs borne by drivers are treated as a separate accounting problem.
Jonathan V. Hall and Alan B. Krueger’s paper “An Analysis of the Labor Market for Uber’s Driver-Partners in the United States,” published in the ILR Review in 2018, is one of the most cited descriptive accounts of ride-hail labour supply in the mid-2010s. Hall was then head of economic research at Uber; Krueger was an academic labour economist. The affiliation is material to how the paper is read: it combines company administrative data with survey responses, and it is open about that partnership.
The study describes who drove, how many hours they worked, how flexible those hours were, and what drivers earned on a gross hourly basis before vehicle costs, fuel and related expenses. Among the headline descriptive results, driver-partners often worked part-time schedules, valued the ability to choose hours, and reported gross hourly earnings that, in the markets and period studied, compared favourably with some local taxi benchmarks on the measures the authors construct. Utilization — the share of online time spent with a passenger — is reported as a separate operational statistic, not as a wage.
That separation matters for anyone reading the paper as a contribution to online monetization research. A gross hourly earnings figure that does not subtract costs is not take-home pay. A utilization rate is not an earnings rate. Hall and Krueger discuss costs as a distinct issue; later work by other researchers has pressed harder on net earnings after expenses and on how those nets vary by city, period and driving intensity. Those later debates do not erase the descriptive value of the original paper, but they do limit how far its gross figures can travel.
Design limits that travel with the figures
The data cover Uber driver-partners in the United States in a specific growth period for the platform. Markets outside the US, later regulatory regimes, and competing ride-hail apps are outside the sample. Survey response patterns and the selection of who remains an active driver-partner can shape averages. The paper is a labour-market description of one firm’s partners, not a randomised evaluation of ride-hail as a category of work.
Critical readings of the paper often focus on three points: the authors’ access to company data, the treatment of costs, and the risk that early-period earnings conditions do not persist. None of those critiques requires discarding the paper. They require keeping the published definition of each figure attached to the figure — which is the rule this ledger applies to every note.
Compared with microtask scrapes such as Hara et al. on Mechanical Turk, Hall and Krueger measure a different labour process: passenger transport with vehicle capital, city-level demand, and an app-mediated matching market. Compared with vacancy indices such as the Online Labour Index, they measure realised hours and earnings for drivers already on the platform, not the flow of new job posts. Placing the three next to each other shows how “side income on the internet” fragments into distinct empirical objects once the measurement design is named.
For editorial purposes, the durable contribution of Hall and Krueger is the clarity of categories: who is counted as a driver-partner, what counts as an hour online, what is reported as gross earnings, and what remains outside the earnings column. Later papers that revise the net picture after costs are doing complementary work; they are not a reason to blur those categories in the original description.