Why survey and administrative counts of gig work diverge

Katharine Abraham, John Haltiwanger, Kristin Sandusky and James Spletzer have published a body of work that places household-survey measures of nontraditional and gig work beside administrative and tax-based measures of the same broad phenomenon. Their comparisons are useful not because they crown one source as correct, but because they isolate the design features that move the measured size of gig activity: what counts as a job, how long the observation window is, and which payments clear a reporting threshold.

Pay Census treats that programme of work as an instrument note. The question is not whether gig work grew or shrank in a particular year. The question is why two carefully constructed totals, aimed at related populations, can disagree without either total being a simple error.

What a household survey typically sees

Survey instruments modelled on labour-force questionnaires ask people about work in a short reference period and often about a main job. Respondents must recognise the activity as employment in the sense the interview uses, recall it within the window, and classify it under categories the form provides. Intermittent platform hours taken beside a primary job, or activity the respondent treats as informal side money rather than a job, are easy to miss even when the questionnaire is well designed.

Surveys also face nonresponse and mode effects. People who are busy with multiple short engagements may be harder to reach; online panels and in-person household samples recruit differently. Abraham and co-authors emphasise that these features are part of the measurement system, not footnotes that can be ignored when a survey share is set next to an administrative share.

What administrative sources typically see

Administrative and tax records observe payments that clear statutory or institutional thresholds: information returns issued when payments exceed a reporting amount, wage records filed by employers, or similar filings. They can cover a full year and capture many small engagements that never become a person’s main job in any given week. They do not, however, observe unpaid search time, work paid below the reporting threshold, or arrangements that never generate the form the researcher is using.

Administrative definitions follow tax and programme rules rather than labour-force concepts. A person may appear as self-employed in tax data while describing themselves differently in a survey, or may receive platform payments that are aggregated, delayed, or labelled in ways that hide the underlying activity. The window is usually annual; the unit is often a payment or a tax identity, not a reference-week job.

Where the totals part company

When Abraham, Haltiwanger, Sandusky and Spletzer set these sources side by side, several mechanisms account for much of the gap. Reference period is one: a week versus a year changes who enters the numerator. Job definition is another: main-job contingency categories versus any payment that triggers an information return. Reporting thresholds pull administrative counts in one direction; recall and classification pull survey counts in another.

Their work also stresses that “gig” is not a single statistical object. Platform-mediated tasks, traditional independent contracting, and contingent employment overlap only partly. Comparing a survey cell built for alternative arrangements with a tax series built for nonemployee compensation can look like a disagreement about the labour market when it is largely a disagreement about which slice each form was cut to measure.

What this note does not claim

This note does not restate a preferred headline size of the gig economy, and it does not treat administrative data as a gold standard that surveys must match. Abraham and colleagues document how each source is produced and where its blind spots sit. A larger administrative total may mean that the administrative form counts more short engagements across a year. A smaller survey total may mean that the survey never asked about secondary jobs in that form.

Why the comparison sits in this series

For anyone reading published counts of platform or online side income, the practical implication is narrow: name the instrument before comparing the number. Survey supplements, tax information returns and bank-transaction studies answer different questions. Abraham, Haltiwanger, Sandusky and Spletzer supply a framework for keeping those questions separate.