Twelve Years When Nobody Counted

The Instrument and What It Measured

The Contingent Worker Supplement was not a standalone survey. It was a periodic add-on to the Current Population Survey, the monthly household survey that the Bureau of Labor Statistics and the Census Bureau have jointly conducted since 1940, and which remains the primary source for US unemployment figures. The Supplement asked a battery of additional questions — of a rotating sample of CPS respondents — designed to identify workers whose jobs lacked an explicit or implicit expectation of long-term continuation. The Bureau first ran it in February 1995, then again in 1997, 1999, 2001, and 2005. Five iterations across ten years, each producing a nationally representative snapshot of what the BLS calls contingent workers: those without a long-term attachment to their employers, including temporary employees, on-call workers, and contract company workers.

The operational definition the Bureau used was deliberately conservative. Under its narrowest measure, a worker counted as contingent only if the job was expected to last less than a year and had already lasted less than a year — a formulation that excluded large numbers of people in economically precarious arrangements who nonetheless held jobs nominally open-ended. The BLS also ran two broader measures that relaxed those conditions, and the spread between them was itself informative: in 1995, the narrowest estimate put the contingent share of the US workforce at just over 2 percent; the broadest approached 4.9 percent. The gap between those numbers was not a rounding error — it was a policy argument encoded in a footnote.

What the Supplement captured, beyond simple headcounts, was a portrait of who contingent workers were. The 2005 iteration, published later that year, confirmed patterns visible in all five waves: contingent workers were disproportionately young, more likely to be Black or Hispanic, less likely to have employer-provided health insurance, and less likely to have a pension plan than workers in traditional arrangements. Independent contractors, by contrast, who were counted separately as a category of alternative arrangement worker, skewed older and were more likely to be white men — a distinction that would matter enormously in later debates about whether non-standard work was a choice or a compulsion.

The Twelve-Year Absence

Then, after the 2005 fieldwork, the Supplement stopped. No official discontinuation announcement was made with the fanfare of a policy decision; it simply was not funded for the 2007 cycle, and then it was not funded again. The BLS cited budgetary constraints, and the Office of Management and Budget's budget requests for subsequent years did not restore the line. By the time the Obama administration was in its second term, more than a decade of structural change in the American labour market — the acceleration of platform-based work, the aftermath of the 2008 financial crisis, the proliferation of app-mediated gig arrangements — had occurred with no systematic federal count of how many workers it had touched.

The absence created a measurement vacuum that others rushed to fill, with predictably variable results. Advocacy organisations, management consultancies, and platform companies all produced figures, and those figures diverged dramatically because they were built on different definitions. The McKinsey Global Institute's 2016 report estimated that between 20 and 30 percent of the working-age population in the United States and Europe engaged in some form of independent work, though its methodology aggregated multiple data sources and its own authors acknowledged the caveats on the first substantive page. Freelancers Union, co-founded by Sara Horowitz in 1995 as Working Today, released annual figures suggesting freelancers made up a rising share of the workforce — estimates that reached 35 percent by some iterations of its commissioned surveys. None of these replaced a federal instrument with a consistent methodology, a randomised sample, and a stable set of definitions applied over time.

The policy consequences were direct. When the California Supreme Court formulated the ABC test in Dynamex Operations West v. Superior Court in 2018, lawmakers and litigants on both sides of the resulting legislative battles had no authoritative count of how many workers in the state actually occupied the contested classification territory. The same was true in Washington, D.C., when Congress debated the PRO Act, and in Brussels, when the European Commission began the negotiations that would eventually produce the Platform Work Directive in 2024. Figures entered policy documents with commissioner attributions rather than agency pedigrees, and opponents of any given regulation could credibly challenge the underlying count while accepting the regulatory question in principle. The absence of a federal instrument did not prevent the debate; it made the debate harder to resolve.

The Return and What It Found

Congress eventually appropriated funds, and the BLS fielded a new Contingent Worker Supplement in May 2017. The twelve-year gap was over, and the 2017 results landed with a surprise that cut against the prevailing narrative: the share of workers in contingent and alternative arrangements had not grown dramatically. The BLS estimated that contingent workers under the narrowest definition accounted for approximately 1.3 percent of total employment — down from the 1995 figures, not up. Independent contractors, the category most often invoked in debates about the gig economy, represented around 6.9 percent of total employment, a figure broadly comparable to 2005.

The finding was contested immediately. Critics argued that the CPS household survey methodology undercounted platform workers because respondents characterised platform-mediated work as a supplement to primary employment, not as their primary job. Researchers Lawrence Katz of Harvard and Alan Krueger, whose 2016 working paper using survey data had suggested that the share of workers in alternative arrangements had grown substantially between 2005 and 2015, disputed whether the Supplement's design was capable of capturing the phenomenon it was meant to measure. The BLS itself acknowledged that the survey asked respondents about their main job, potentially missing those who patched together multiple contingent engagements.

This methodological argument — about whether the instrument was measuring the right thing — was only possible because the instrument had returned. For twelve years, there had been nothing to argue with. Policy had proceeded on the basis of private-sector estimates that carried their commissioners' interests, advocacy figures that reflected their organisations' purposes, and occasional academic extrapolations from administrative data not designed for the task. The Supplement's absence had not kept independent work out of the legislative agenda; it had simply removed the government's own voice from the factual foundation of that agenda.

The 2017 return also underscored what the 2005 data had already shown: the categories inside the Supplement matter as much as the Supplement itself. Contingent workers, independent contractors, on-call workers, temporary agency staff, and contract company workers each describe a different relationship to work and a different legal and economic exposure. Collapsing them into a single headline figure — as non-governmental estimates routinely did during the twelve missing years — produced a number that was simultaneously true and misleading, large enough to motivate policy action and imprecise enough to obstruct it. A measuring instrument's design is never politically neutral, and its discontinuation, as the decade after 2005 demonstrated, is a political act whether or not it is intended as one.