What the Online Labour Index measures — and what it leaves out

Otto Kässi and Vili Lehdonvirta introduced the Online Labour Index (OLI) in work associated with the Oxford Internet Institute, with a widely cited presentation of the project appearing in 2018 (“Online labour index: Measuring the online gig economy for policy and research,” Technological Forecasting and Social Change). The index aggregates new job vacancies posted on a set of large English-language online labour platforms and normalises them into a time series that can be broken down by occupation and by employer country.

The measurement object is demand as expressed in new posts. When the index rises, more vacancies are appearing on the covered platforms. When it falls, fewer new vacancies are appearing. That is a useful signal for researchers and policymakers who need something more structured than anecdote about the online contract labour market. It is also easy to misread. A vacancy is not a filled job. A filled job is not a completed hour. A completed hour is not a reported wage. The OLI sits at the first step of that chain.

Platform coverage is another hard boundary. The original construction focused on a small number of major English-language platforms that post remote project and task vacancies at scale. Local-language platforms, app-based local services, microtask markets structured differently from those platforms, and informal arrangements arranged outside vacancy boards are outside the index by design. Geographic breakdowns in the OLI refer primarily to where employers are located, which is not the same as where workers reside.

Why demand is not earnings

Discussions of online side income often treat “more work online” as if demand, utilisation and pay moved together. The OLI is a reminder that they need not. An increase in posted vacancies can coincide with falling offered rates, longer queues of applicants, or unpaid search time that never appears in vacancy counts. Conversely, stable vacancy totals can hide shifts in occupation mix — for example toward software development and away from data entry — that change what kinds of skills are being sought.

Kässi and Lehdonvirta are explicit that the index is a labour-demand measure built from platform vacancy data. Later extensions and related datasets have expanded coverage and occupation taxonomies; those extensions inherit the same conceptual limit unless they add separate earnings or hours modules. For this ledger, the methodological takeaway is simple: cite the OLI for what it measures. Do not treat a vacancy index as evidence about individual earnings distributions of the kind reported by Hara et al. for Mechanical Turk, or about utilisation and gross hourly earnings of the kind described by Hall and Krueger for Uber driver-partners.

Kept in that frame, the Online Labour Index is a precise tool. It shows how online contract demand moved across occupations and employer countries on the platforms it covers. It does not settle debates about how much people earn from platform work, how many hours they complete, or how online monetization appears in household income. Those questions require other designs — and other notes.

An editorial habit that follows from the OLI is to ask, of any chart about “online work,” whether the underlying series counts posts, contracts, hours or pay. When that question cannot be answered from the source, the chart does not belong in this ledger. When it can, the series can be cited with its unit attached — which is what Kässi and Lehdonvirta make possible for vacancy demand.