How we measure: the informal economy correction

Countries corrected

145 of 217

128 income, 130 wealth, 113 of them both

Ceiling on a shift

2.00x

it binds on 54 corrected distributions

What moves

The level

spread and tail stay as published

What the correction does

Share of the distance between the measured median and the one output implies that the correction closes. The ten most populous countries where it is in force on income.

In Nigeria the correction closes 92% of the distance between the measured median and the one its output implies. In the United States it closes 6%.

72countries the correction leaves as published

2.00xthe ceiling on any single shift

Where a country's own records measure less activity than its output implies, we raise the level of its distribution toward the implied figure and leave the spread and the tail as published. The correction is in force on 145 of 217 countries. In India, where the informal share of output is 50.5%, it lifts the median income from $1,422 to $2,503; in the United States, at 8.3%, from $50,706 to $51,703.

A worked example

Step	India	United States
Informal share	50.5%	8.3%
Measured median	$1,422	$50,706
Implied median	$2,730	$67,579
Gap	1.92x	1.33x
Blend weight	0.83	0.06
Factor	1.76x	1.02x
Corrected median	$2,503	$51,703

Income per adult, base year 2026, converted to US dollars at the rate held with each record. Informal share is informal output as a percent of official GDP. The implied median is output per person times the trusted ratio. The share carried for India is dated 2024, for the United States 2020.

Questions people ask

What is the informal economy correction?

Tax records and household surveys miss cash work and self-employment. Where a country's measured median sits below what its output per person implies, we raise the level of its distribution toward the implied figure, in proportion to how large its informal economy is. The spread and the tail are published as measured.

How many countries does it change?

It is in force on 145 of the 217 countries published here: 128 income distributions, 130 wealth distributions, 113 of them both. The remaining 72 countries are published as measured.

How is the size of the correction decided?

Two numbers. The gap is the output-implied median divided by the measured one, capped at 2.00x. The blend weight rises from 0 at an informal share of 5% to 1 at 60%. The factor is 1 plus the weight times the gap above 1, so a country whose own records are trusted moves little even when its gap is wide.

Can the correction lower a published figure?

It raises only. Where a country's measured median already sits above the output-implied one, which is the pattern in states with large expatriate workforces, the figure is published as measured. A gap of 1.05x or less is left alone as well.

Where does the informal-economy figure come from?

The World Bank Informal Economy Database, which expresses informal output as a percent of official GDP, taken together with the International Labour Organization's vulnerable-employment share scaled to 70 percent. Whichever of the two is larger is used. The shares carried on corrected countries are dated 2020 and 2024.

Method and sources

World Bank Informal Economy DatabaseInformal output as a percent of official GDP. Elgin, Kose, Ohnsorge and Yu.
International Labour OrganizationVulnerable employment share, scaled to 70 percent and used where it is the larger of the two readings.
World Bank World Development IndicatorsGDP per capita in current US dollars, the anchor for the implied median.
World Inequality DatabaseThe measured income and wealth distributions the correction is applied to.

How the trusted ratio is set

Countries with an informal share of 12% or less and GDP per capita of $15,000 or more seed a trusted ratio: the median of measured median divided by output per person across that group. Applying that ratio to any country's output gives the median its measurement would show if its records were as complete as theirs.

How the factor is applied

As a level shift and nothing else. The log-normal centre moves by log of the factor, the Pareto lower bound scales by it, and every threshold and percentile scales by it. The log-normal spread and the Pareto exponent are untouched, so the ratio between any two percentiles in a country is identical before and after.

Where a later step replaces this one

On 32 country dimensions the correction is computed and then superseded. 19 are overwritten by a GDP anchor, used where the measured median still sat more than three times away from output in either direction, and 13 wealth distributions use a published UBS median instead. Those records carry the later figure and are not counted among the 145.

What these figures do not cover

Informal output is itself an estimate. The World Bank series models activity that is by definition not directly measured, and two model families that disagree are averaged rather than reconciled.
The correction is capped at 2.00x, so a country whose measured median sits further than that below its output-implied figure is still published below it. That cap binds on 54 of the corrected distributions.
A country with neither an informal-output figure nor a vulnerable-employment figure is published exactly as measured, which is one reason 72 of the 217 countries here are untouched.
The correction changes the level of a distribution, so it cannot change any inequality reading. A country's percentile ratios are the same before and after.

Last updated 2026-07-25.

See where you land Compare a net worth

Tesseract Stock Agent is a professional-grade AI research agent with one job: analyze any stock the way institutional equity research does. The same evidence standard that built this page, pointed at your portfolio.

See Tesseract Stock Agent