Abraham and colleagues: the survey–administrative gap in nontraditional work

Katharine Abraham, John Haltiwanger, Kristin Sandusky and James Spletzer have published a line of research that compares how household surveys and administrative tax records portray nontraditional and electronically mediated work. Pay Brief treats that literature as a map of measurement seams: places where legal form, tax filing and survey wording diverge, so that the same economic activity can be counted, misclassified or missed depending on the instrument.

Two ledgers, different objects

Household surveys such as the CPS ask respondents to describe jobs, arrangements and, in supplements, contingent status. Administrative sources—notably information returns and self-employment income reported to tax authorities—observe filings associated with payers and payees. Abraham and coauthors emphasize that these systems were not designed as mirrors. A surge in 1099-type information returns can signal growth in contracting relationships without implying an identical surge in survey-reported independent contracting, because respondents may not use the same labels, may omit small amounts, or may fold platform activity into a wage job narrative.

Conversely, survey questions about “gig” or online intermediary work can surface activity that never crosses a reporting threshold on a tax form, or that is reported under a different schedule. The gap is therefore two-sided: administrative data can overstate or understate the concept a journalist attaches to “gig work,” and surveys can undercount intermittent online side income that tax systems partially see.

Form versus activity

A recurring warning in this research is the confusion of legal or tax form with economic activity. Self-employment on a tax return is not synonymous with platform-mediated task work; it includes offline trades, partnerships and other arrangements. Platform firms may classify payees as independent contractors under commercial contracts, while labor-force surveys may code the same person as a wage employee in a different primary job. Abraham, Haltiwanger, Sandusky and Spletzer stress careful taxonomy: without stating whether the unit is persons, jobs, spells or dollars, headline comparisons across sources become rhetoric rather than measurement.

Undercount and overcount risks

Undercount in surveys arises from recall error, stigma or ambiguity about whether a few hours of app-mediated work “counts” as a job. Proxy respondents amplify that risk. Undercount in tax data arises when payers are not required to file information returns below thresholds, when payees fail to report, or when receipts are mischaracterized. Overcount appears when analysts interpret all nonemployee compensation as platform work, or when year-over-year growth in filings is read as labor-market transformation without adjusting for reporting-rule changes and firm compliance.

The authors’ comparative approach does not crown one source as truth. It shows that reconciling series requires explicit crosswalks: which survey question aligns with which tax concept, for which years, under which filing rules. Absent that crosswalk, debates about whether household surveys “miss the gig economy” often talk past the administrative evidence.

Implications for reading side-income claims

For an editorial desk interested in online side income, the Abraham–Haltiwanger–Sandusky–Spletzer agenda implies three reading rules. Name the instrument. State whether the object is persons, jobs or dollars. Treat disagreement between survey and tax series as a methodological finding, not as a license to invent a third number. Pay Brief applies those rules when juxtaposing BLS supplements with administrative studies and does not translate measurement gaps into advice about personal earnings.

← Back to index