Earn Index

Notes on published research about platform work — and on the estimates that were later revised

The revision sits beside the original figure whenever the researchers or the statistical agency published one.

Instrument Survey · Tax · Agency Status Corrected · Revised · Compared Units Kept apart

Subject

Corrections are the record, not the footnote

Headline numbers about online earnings travel further than the papers they came from. A share of the workforce, an hourly figure, a growth rate: each is produced by a particular instrument, over a particular period, for a particular population, and each carries the limits of that instrument.

Some of the most cited estimates in this field were subsequently corrected by the people who produced them. A survey question was misread by respondents. A comparison across two datasets turned out to be inconsistent. A tax record captured a different population than a household survey did.

Those corrections are not footnotes here. They are the subject. Each note states what was measured, what the first published figure was, what the revision changed, and what the authors said their result does not cover.

Three measures, and what happened to them

Units kept apart

Described as the cited authors and agencies published them. Figures below are not averaged together and are not treated as interchangeable measures of the same thing.

Reference index of notes
Source	Instrument	Status
US Bureau of Labor Statistics	Contingent Worker Supplement, electronically mediated work	Corrected by agency
Katz & Krueger	Alternative work arrangements, RAND survey vs. household survey	Revised by authors
Collins, Garin, Jackson, Koustas & Payne	Federal tax records, two decades of filings	Different instrument

The index

Three notes on record

Each entry sets out the question, the instrument, the published result and the boundary the authors placed around it.

Entry 01Instrument Household survey supplementStatus Corrected 2018

A federal statistic that its own agency withdrew

The US Bureau of Labor Statistics published estimates of electronically mediated work from the May 2017 Contingent Worker Supplement, then determined that a substantial number of respondents had been misclassified, and reissued lower figures with an explanation of the error.

Entry 02Instrument Survey comparisonStatus Revised by authors

When two economists revised their own widely cited estimate

Lawrence Katz and Alan Krueger reported a sharp rise in alternative work arrangements, then published a reassessment concluding that the increase they had measured was considerably smaller once their survey was set against the government's own supplement.

Entry 03Instrument Administrative tax recordsStatus Different population

What tax filings show that household surveys miss

Collins, Garin, Jackson, Koustas and Payne read two decades of federal tax returns and found platform work concentrated among filers with modest and often intermittent income from it — a picture drawn by a different instrument than the surveys it is frequently compared against.

Editorial standards

Four rules that govern every page

Named sources only. Every figure on this site is attributed to named researchers or a named institutional publication, with the instrument and the period attached. No statistic appears here without a traceable source.
No presumption about the reader. Every page is written in the third person about research populations. No headline, sentence or graphic asserts or assumes anything about the situation, finances, occupation or intentions of the person reading it.
Limits travel with the figure. Where authors state that a result does not extend to a population, industry or period, that boundary is reproduced in the note rather than left to inference.
No accounts of individual results. This publication does not publish testimonials, customer reviews, endorsements, success stories or before-and-after accounts, in any section, and does not collect them.

Boundaries

What Earn Index is not

Set out directly, so that the edges of this work are not left to inference.

Not financial advice: Nothing here is professional, financial, investment, tax, legal, employment or career advice, and nothing here substitutes for a qualified professional who has examined the particulars of a situation.
Not an assessment: This publication does not evaluate, diagnose or assess anyone's individual circumstances, and offers no consultation, coaching or mentorship.
Not a claim of results: No article states or implies an outcome for anyone who reads it. A measured average or median in a study population is a statement about that population and nothing else.
Not testimony: No testimonials, customer reviews, endorsements, success stories or before-and-after accounts appear anywhere on this site, and none are solicited.
Not a support service: This is a publication, not a crisis line or helpline. Correspondence is read during business hours and is not monitored continuously.
Not for minors: The site and anything associated with it are intended for readers aged 18 or over.

Research described here concerns samples, populations and averages measured under stated conditions in stated periods. A finding about a population is not a prediction about an individual, and is not presented as one anywhere on this site.

Questions

Answered in advance

What is published here: Editorial notes on surveys, administrative datasets and measurement projects concerning platform labour, task pay and online labour demand. Each note names the researchers or the agency, describes the instrument and the population, reports the figures as published, and records the boundaries placed on the conclusions.
Whether reading costs anything: It does not. The index on this site is open, requires no account and no payment. Where a paid compilation is offered separately, it is governed by the Terms of Access and the Refund Policy linked at the foot of every page.
How refunds work: Any paid item carries a 30-day refund window counted from the purchase date, with no reason required. Requests go to the correspondence address in writing. The Refund Policy sets out the timings and the statutory rights it does not replace.
Whether consultations are available: They are not. There is no advisory service, no one-to-one guidance and no review of anyone's situation. Correspondence covers questions about published notes, corrections, and administrative matters.
Why there are no testimonials: Because an account of one person's result is evidence about one person, and this publication reports study populations. The policy is absolute: none are published, and none are collected.
What data is recorded: There is no account and no comment form. Delivery logs, any audience measurement in use, and correspondence are described in the Privacy Notice, along with the legal bases, retention periods and the rights available under the LGPD and the GDPR.

Correspondence

One address, answered in writing

Earn Index is reachable at a single email address, in English or Portuguese. Replies are sent within two business days, Monday to Friday, 09:00–18:00 Brasília time. There is no telephone line and no office open to visitors.

[email protected]

Corrections are handled under the policy on the Method page. Requests concerning personal data are handled under the Privacy Notice.