Seventeen Percent, Inside the Same Ethnicity

The skin lightening industry sells an outcome, not a color. Better job. Better marriage. Better treatment.

So the honest question is not whether the creams work. It is whether the thing they promise is real, and what the measurements actually say — including where they say nothing, and where they say the opposite.

The wage finding, and why the controls matter

Joni Hersch used the New Immigrant Survey, 2003, a sample of legal permanent residents. Skin color was rated by the interviewer against an eleven-point visual chart showing hands of varying darkness. Not self-reported, and rated on the hand rather than the face.

In the specification controlling for pre-US characteristics, ethnicity, race, and country of origin: a coefficient of minus 0.017, significant at p below 0.01. Roughly a 1.7 percent wage reduction per point of darkness.

Her summary: immigrants with the lightest skin color earn on average 17 percent more than comparable immigrants with the darkest. Moving from the tenth to the ninetieth percentile of the skin color distribution reduced wages by 6.8 to 8.5 percent.

For scale, in the same analysis, height was worth about 2 percent per inch above the gender-specific average.

The controls are the entire point. This is not national origin wearing a disguise. It is a gradient within ethnicity, race and country of origin — people from the same place, doing comparable work, paid differently by shade.

The canonical study of US-born Black Americans exists, is heavily cited, and is closed-access with no abstract I could retrieve. I am not going to reproduce numbers from it that I have only seen second-hand. Same for the marriage literature and the sentencing literature: the papers exist, the direction is consistent, and the specific effect sizes that circulate — a percentage shorter sentence here, a doubled likelihood there — could not be checked against the originals in this pass. One of those findings has a boundary condition about victim race that gets dropped in nearly every retelling, which is exactly why I would want to read it before repeating it.

What is verifiable, straight from the abstract of Ellis Monk’s 2015 paper on skin tone, discrimination and health among African Americans: intraracial health differences related to skin tone and discrimination often rival or even exceed disparities between Black and white people as a whole.

That sentence is the one to hold onto. A gradient inside a group that is as large as the gap between groups.

One methodological caution the field raises about itself. Monk also found that self-reported skin tone predicted perceived discrimination better than interviewer-rated skin tone. And two separate papers have questioned the reliability of interviewer observation of skin tone, including across racial lines. A great deal of this literature rests on one person’s eyeball estimate of another person’s color, and that instrument is shakier than the statistics built on it.

The experiment that came back null

Vijaya and Bhullar ran an experimental survey in India: participants evaluated job candidates using identical résumés paired with photographs digitally manipulated to vary skin tone.

Their result, verbatim: we did not find a statistically significant bias in favor of résumés paired with lighter-skinned candidates.

The Indian fairness cream market sells an employment-success claim. The one published experimental test of that claim in India returned null.

That is not a refutation of colorism, and the authors do not present it as one — they note the motivation for the study was the lack of skin-tone-specific data globally. But it belongs in any sober treatment, and it is the sort of finding that usually gets left out of coverage on both sides.

How many people, and where

The pooled global figure comes from a meta-analysis of 68 studies and 67,665 participants: lifetime prevalence of skin bleaching, 27.7 percent.

Heterogeneity, I-squared, 99.6.

Print the second number beside the first or the first is misleading. Ninety-nine point six percent heterogeneity means the studies are not measuring comparable things, and a global rate assembled from them is a statistical artifact rather than a description of the world.

The country-level surveys are more useful, and they are wide-ranging.

And in the pooled analysis the figure for men — 28.0 percent — came out slightly above the overall pooled figure. This is not a women-only practice in the data, however it is marketed.

Countries I could not source at all: Senegal, Pakistan, the Philippines, South Korea, Jamaica. No population prevalence survey located for any of them. Those gaps are part of the picture, and any global map that fills them in is filling them in from somewhere other than a survey.

The mercury finding that cuts against the assumption

This is the best piece of hard chemistry in the subject and almost nobody reports it correctly.

Pramanik and colleagues measured mercury in skin care products sold in India. Median concentrations, in micrograms per kilogram:

Body lotions, 0.11. Moisturizing lotions, 0.22. Blemish balms, 0.67. Bleaches, 0.16. Domestically manufactured fairness creams, 0.13.

Imported fairness creams: 9.24 million.

The regulatory limit is 1 part per million, which is 1,000 micrograms per kilogram. The imported creams came in at roughly 9,240 parts per million — about nine thousand times the legal limit, and seven orders of magnitude above the domestically made products.

Dermal exposure hazard quotients above 100 in all cases.

Indian-manufactured fairness creams were essentially mercury-free. The contamination is an import-channel problem, not something indigenous to that market — which is the opposite of the way this usually gets written about.

The complication that follows long-term hydroquinone use is exogenous ochronosis, a blue-black pigmentation that is clinically and histologically similar to a metabolic disease, and that the literature describes with unusual bluntness as very difficult to treat with disappointing treatment options.

One correction on attribution while we are here. The WHO mercury and health fact sheet states that skin lightening is often associated with exposure to mercury salts, and that mercury-containing lightening products are hazardous and banned in many countries. It gives no country-level prevalence figures and no numeric limit for mercury in cosmetics. If a prevalence percentage is attributed to WHO, it came from a different document, and somebody should name it.

What has been tried, and what has been measured

Here is the finding that surprised me most.

I went looking for intervention, campaign, health promotion, education, ban and policy studies on skin lightening. Nothing I found is a controlled trial of an intervention to reduce use.

The closest things are a qualitative focus group of forty Cameroonian girls proposing educational strategies with no outcome measurement, and a philosophy paper arguing for a relational ethic of polycentric governance.

No published, evaluated intervention has ever been shown to reduce skin lightening product use. The entire field is descriptive.

The regulatory side is similarly thin, at least as far as I could verify. India is the obvious test case and I could not confirm the two things everybody cites. I could not obtain a document titled as the advertising standards council’s fairness-product guidelines — three retrieval attempts, no clause number, no effective date. What the main code does contain is a general provision requiring special care in advertising to people with a real or perceived inadequacy of any physical attribute, which would cover fairness claims by implication and is not a category-specific rule.

And I could not verify the 2020 amendment to the advertising law that is routinely described as banning fairness claims — not its contents, not the penalties in circulation, and critically not whether it was ever enacted. The likeliest position is that it was a draft released for comment and never passed. That is my inference, not a finding, and I would rather say so than print two prison terms and two fines as though they were law.

No enforcement data. No market size figure with a peer-reviewed or official source — every dollar figure in circulation traces to commercial research firms whose methods are undisclosed.

The nearest thing to a measurement of the marketing side is a content analysis of the hundred most-viewed TikTok videos under a skin-lightening hashtag. Seventy-nine were analysable. Forty-three percent of creators were presumably South Asian, 30 percent African American or Black. Most videos provided no scientific evidence for their claims.

The gradient, the null, and the mercury

There is a measured wage gradient inside ethnic groups, and it is substantial. There is a measured health gradient inside Black Americans that its author describes as rivaling the Black-white gap. Those are real and they are the reason the market exists.

There is also a null experiment in the country whose industry sells the employment claim hardest, an instrument problem running under a lot of the measurement, and a pooled prevalence figure with 99.6 percent heterogeneity that should not be quoted alone.

And there is a documented chemical harm with a clear channel — imported creams at nine thousand times the mercury limit while domestic ones were clean — that would be addressable by customs enforcement, which is unglamorous and testable and has not been evaluated either.

Sixty-five percent of schoolgirls in one Ghanaian survey, 71 percent of them asked to start by someone they know. Whatever is being sold, it is not being sold by advertising alone, and nobody has ever run a study on how to interrupt it.

This article is educational and is not a substitute for care from someone who can look at your skin. Anything new, changing, painful or spreading should be assessed rather than researched.