Project Enlightenment · Dossier · Science & Power

Science & Power

How the market and the state bend the evidence — result falsification and the asymmetry of the lie.

The popular thesis goes: science and capitalism are incompatible. That is too crude — and too narrow. The rupture always appears in the same place: wherever power makes the truth expensive. Under the market the lever is profit; under ideological states it is doctrine and plan. Then, reliably, the evidence bends.

A large balance scale: one pan holds a stack of papers, the other is dragged down by a giant hand dropping gold coins. — Plate · The Bent Scale Where power makes the truth expensive, the evidence *bends.*

Kennzahl: 36%; of psychology findings could be replicated
Kennzahl: ~2×; odds that sponsored studies come out “positive”
Kennzahl: 10,000+; retracted papers in 2023 alone — a record
Kennzahl: ~40%; profit margin of the largest science publisher

The Findings · I–III

Three measurements, one pattern

Before we turn to individual offenders: the problem is systemic and measurable. Three independent lines of research show the same thing — not every published finding holds, and money shifts the direction of the results.

I · The Replication Crisis

What happens when you simply run published studies again.

Kennzahl: 36%; of the replicated psychology studies stayed significant — against 97% significant originals. Effects half as large.
Kennzahl: 11%; of the “landmark” cancer studies Amgen could reproduce (6 of 53).
Kennzahl: ~25%; validatable preclinical studies at Bayer (“Believe it or not”).
Kennzahl: 70%; of researchers failed to reproduce others' experiments, >50% their own (Nature survey).

As early as 2005 Ioannidis published his most-cited paper, titled “Why Most Published Research Findings Are False.” The crisis touches every field — from psychology to oncology.

II · The Funding Effect

Cochrane review (Lundh et al. 2017): 75 papers, 4,583 studies. Manufacturer-sponsored studies come out systematically more favorable — a bias the usual quality checks fail to catch.

Favorable conclusion odds ratio · Lundh 20172.69×

Favorable result odds ratio · Lundh 20172.05×

Earlier estimate Bekelman 20033.60×

Earlier estimate Lexchin 20034.05×

The ironic part: industry studies sometimes had the lower risk of bias in blinding. So the distortion does not sit in the obvious craft, but in the subtle — in the framing of the question, the study design, selective reporting and the “spin” of the conclusion.

III · Publication Bias

Turner et al., NEJM 2008 — antidepressant approval trials checked against the FDA files.

Kennzahl: 31%; of 74 FDA-registered trials were never published.
Kennzahl: 94%; of the studies looked positive in the published literature …
Kennzahl: 51%; … according to the FDA files, only this many actually were.
Kennzahl: +32%; by this much the apparent efficacy was inflated.

Publish only the positive studies and a drug looks more effective than it is — without falsifying a single number. Omission is enough.

The Files

Top 10 Fraud Cases

The market first: ordered not by fame, but by real harm — deaths, misguided policy, lost years. Every case follows the same pattern: a claim was cheap to produce; the exposure was expensive and came late. The state half follows directly below.

01 · from the 1950s · Public Health · Manufactured Doubt

The Tobacco Industry

Internal studies proved the cancer risk early. Outwardly, paid scientists systematically manufactured doubt. An internal memo (R.J. Reynolds, 1969): “Doubt is our product.” The goal was never to be right — but to generate controversy.

Auswirkung Decades of delay in regulation and compensation; millions of avoidable deaths. The playbook was later reused almost unchanged for acid rain, the ozone hole and climate change (Oreskes/Conway, “Merchants of Doubt”).

02 · from 1996 · Pharma Marketing · Opioid Crisis

OxyContin · Purdue Pharma

Sales pitch: “Fewer than 1% of patients become addicted.” This rested, among other things, on a 110-word letter to the editor from 1980 (Porter & Jick) that proved no such thing — written 16 years before the drug's launch.

Auswirkung A co-trigger of the US opioid epidemic — over 500,000 deaths in two decades. Guilty pleas in 2007 ($634.5 million) and 2020 (settlement framework up to $8.3 billion).

03 · 1999–2004 · Pharma · Data Suppression

Vioxx · Merck

The cardiovascular risk was concealed. In the VIGOR trial, heart-attack data were deleted before the NEJM submission; in 16 of 20 studies a Merck employee was initially listed as first author (ghostwriting).

Auswirkung An estimated 88,000–140,000 additional cases of heart disease, of which roughly 38,000–55,000 fatal. Withdrawn from the market on September 30, 2004; legal liability estimated at up to ~$50 billion.

04 · 1965–67 · Nutrition Science · Covert Funding

The Sugar Industry

The Sugar Research Foundation paid Harvard researchers (~$6,500, ≈ $48,000 today) for an NEJM review that blamed fat and exonerated sugar. The funding was not disclosed; a draft was submitted to the sponsor in advance.

Auswirkung Shaped the “fat, not sugar” orientation of nutrition research and policy for decades — with consequences reaching into today's obesity and diabetes debates.

05 · 1998 · Vaccines · Falsification

Wakefield · MMR & Autism

A fabricated Lancet study of just 12 children linked the MMR vaccine to autism. Wakefield received >£600,000 from a litigating lawyer; partly retracted in 2004, fully in 2010, medical license revoked.

Auswirkung The vaccination rate in England 1996–2002 fell from 91.8% to ~81% (in London <60%); the return of measles. The debunked myth persists 25+ years later — the textbook case of the asymmetry (see below).

06 · from 2011 · Surgery / Stem Cells · Deadly

Macchiarini · Karolinska

Synthetic, stem-cell-coated tracheas were described in the Lancet as a “success” — complications were concealed, one finding (“four-month check-up without complications”) demonstrably falsified.

Auswirkung 7 of 8 transplant patients died. Convicted in Sweden in 2023 (30 months in prison for assault); several Lancet retractions.

07 · 2003–2018 · MedTech · Investor Fraud

Theranos · Elizabeth Holmes

A “single-drop” blood test that never worked reliably. Patients received false results (among them a false-positive HIV test, a wrongly triggered change of medication).

Auswirkung Valuation at times ~$9–11 billion. Holmes: 11 years 3 months in prison, Balwani ~13 years; $452 million in restitution. Conviction upheld on appeal in 2025.

08 · 2004–05 · Stem Cell Research · Fabrication

Hwang Woo-suk

Allegedly the world's first cloned human embryonic stem cell lines (Science). The data were fabricated, the egg cells procured unethically.

Auswirkung Both Science articles retracted; convicted in South Korea in 2009 of embezzlement and violations of the bioethics law. An entire field of research was misdirected for years.

09 · 2020 · COVID-19 · Phantom Data

Surgisphere

A supposed global patient database underpinned two high-profile studies (Lancet on hydroxychloroquine, NEJM). The data could never be independently verified — the company would not disclose them.

Auswirkung Both studies retracted within weeks; clinical trial arms (including at the WHO) were temporarily halted — in the middle of a pandemic this distorted research worldwide.

10 · 2023 · Neuroscience · Failure to Correct

Tessier-Lavigne · Stanford

An investigation confirmed manipulated data in labs he led. He himself was not accused of falsification — but he failed to correct the scientific record for years, though the problems were known.

Auswirkung Resignation as president of Stanford University (August 31, 2023); several retractions and corrections, including a high-profile Alzheimer's study.

The Other Half

Science under the State

It is not “capitalism.” Swap profit for ideology and plan — and the same machine keeps running. Only the sanction against dissenters gets harsher: not lawsuits and loss of funding, but the Gulag, professional bans, death. Where a power makes the truth expensive, the evidence bends — no matter the color of the flag.

01 · 1948 · Genetics · Ideology over Evidence

Lysenko · Soviet Union

Trofim Lysenko declared Western genetics a “bourgeois pseudoscience” — personally backed by Stalin. In 1948 genetics was effectively banned in the USSR, textbooks rewritten, experimental stocks destroyed.

Auswirkung The geneticist Nikolai Vavilov died in the camps in 1943 (of starvation); over 3,000 biologists were dismissed, imprisoned or shot. Soviet biology and agriculture were set back until the mid-1960s — famines made worse.

02 · 1958–62 · Agriculture/Ecology · Doctrine over Warning

Mao · War on the Sparrows

In the “Four Pests” campaign ~2 billion sparrows were killed — against the explicit warning of the biologist Zhu Xi. Added to this were Lysenko methods imported from the USSR (extreme close planting).

Auswirkung Without sparrows, pests like locusts exploded; the harvests collapsed. The sparrow campaign alone accounts for ~20% of the crop collapse and ~2 million direct deaths — part of the Great Famine with 15–55 million dead. Cadres reported invented successes up the chain.

03 · 1930s · Physics · “Race” over Truth

“Deutsche Physik” · Nazi State

The Nobel laureates Philipp Lenard and Johannes Stark branded Einstein's relativity as “Jewish physics” and sought to impose an “Aryan” physics; Jewish researchers were forced out of the universities.

Auswirkung The emigration of the world's best minds (Einstein and many others) and the permanent loss of Germany's leading position in physics. The ideology only failed when, of all things, the “Jewish” quantum and nuclear physics became crucial to the war.

04 · 1968–89 · Sports Medicine · Prestige over People

GDR State Doping · Plan 14.25

Science in the service of state prestige: under “State Plan Theme 14.25,” ~9,000 athletes — including children from about age 10 — were secretly given Oral-Turinabol, overseen by up to ~1,500 doctors and scientists. Almost no athlete ever failed an official test.

Auswirkung Severe long-term damage: liver damage, tumors, heart disease, infertility, birth defects in the children of those affected. A state-organized, medically supervised human experiment.

05 · 1976–83 · Economics/Statistics · The Plan Forces the Lie

Plan Statistics · Pripiski & Cotton

Where the quota must be “met,” numbers get invented — pripiski. The biggest case: the Uzbek cotton affair. Billions of rubles flowed for cotton that never existed; exposed only by spy satellites that showed empty fields.

Auswirkung Roughly 4 billion rubles in damages; one of the largest purges after Stalin — about 58,000 officials affected, ~20,000 criminally prosecuted. Statistics as an instrument of control were destroyed.

06 · Present Day · Research System · Bureaucratic Incentive

China Today · the Paper-Mill Industry

Promotions and bonuses hinge on publication counts — so people buy them. A whole industry sells fake studies and authorships. Not a profit motive in the market sense, but bureaucratic plan pressure.

Auswirkung A co-driver of the more than 10,000 retractions of 2023; the highest retraction rates are found in, among others, Saudi Arabia, Pakistan, Russia and China. The same Goodhart mechanism as “publish or perish” in the West — just on a different beat.

The Synthesis

One Mechanism, Two Levers

The market and the state look like opposites. In their relation to the truth they are the same machine with two levers — one buys, the other coerces.

Lever: Profit

The Market

What bendsStudies, approvals, dietary guidelines — in favor of the product.

Sanction against dissentersLoss of funding, end of career, lawsuits, character assassination.

Camouflage“More research needed,” manufactured doubt, “spin.”

ExamplesTobacco · Vioxx · OxyContin · Sugar.

Lever: Ideology & Plan

The State

What bendsWhole disciplines, statistics, textbooks — in favor of the doctrine.

Sanction against dissentersProfessional bans, camps, execution.

Camouflage“Bourgeois pseudoscience,” plan fulfillment, patriotism.

ExamplesLysenko · Sparrows · “Deutsche Physik” · Doping.

The Lever

The Asymmetry of the Lie

Why does falsification pay off at all? Because making a claim is cheap and exposing it is expensive. This gap is the real mechanism behind every case above.

An off-balance seesaw: a tiny figure whispers on the short, heavy end while a whole crowd of speakers on the long, raised end pushes down in vain. — Plate · The Asymmetry One claim outweighs a crowd of rebuttals — the *gap* never closes.

Three? Ten? The direction is what counts.

The claim that exposure costs three times the effort of lying is really the conservative lower bound. The best-known formulation is Brandolini's law (Alberto Brandolini, 2013): the energy needed to refute bullshit is an order of magnitude higher than the energy to produce it — roughly ten times. Brandolini himself explicitly calls it an observational rule of thumb, not a measured law of nature.

Kennzahl: 3×; conservative lower bound — the cautious reading
Kennzahl: 10×; Brandolini's law — “one order of magnitude,” 2013
Kennzahl: ∞; Gish gallop — ten lies a minute, uncatchable

Why the gap widens

Asymmetric effort. A claim needs one sentence. A refutation must gather evidence, explain nuances and anticipate counter-objections — often hours to days.
The Gish gallop. Fire off ten false claims in a minute and you tie up the fact-checker for hours. Volume beats substance.
The continued-influence effect. Even after a clear correction, the original misinformation keeps working in memory — a correction never fully “overwrites” it.
The spread gradient. Truth is slower than the lie — and that has been measured.

The evidence: lies travel faster

Vosoughi, Roy & Aral, Science 2018 (MIT) — the largest study on this to date: ~126,000 stories, shared 4.5 million times by ~3 million people, 2006–2017. False reports spread farther, faster, deeper and wider than the truth in every category.

Kennzahl: 6×; longer the truth takes to reach 1,500 people
Kennzahl: 20×; longer to jump across ten levels of resharing
Kennzahl: 70%; more often falsehood is retweeted than truth

Visible in this dossier's own lab

Wakefield: one fabricated study in 1998 — 12 years to full retraction (2010). The myth outlives it by decades.
Vioxx: deleting data took a single move. Bringing it to light took years, independent re-analyses and an FDA whistleblower.

That is the economic punchline: falsification is cheap, exposure is expensive — and the harm is externalized. It is exactly this asymmetry that makes result falsification rational, as long as no one charges the true costs to the one who caused them.

The Economic Layer

The Publishers: Paying Three Times

Above the individual studies sits an industry that extracts profits from science outside the logic of the market. The pattern is called “triple-pay”:

The public pays for the research through taxes and grants.
Scientists supply the texts and peer review mostly for free.
Universities buy the result back at a premium — and publicly funded research ends up behind paywalls.

A scholar with a satchel pays a coin at each of three successive turnstiles before one enormous locked archive door with a golden keyhole. — Plate · Paying Three Times Fund it, give it away, buy it back — three tolls at the *same* door.

Kennzahl: ~40%; profit margin at Elsevier/RELX (2023) — on par with Apple or Google.
Kennzahl: 3.6 bn; profit of the parent company RELX, 2023 (in $).
Kennzahl: ~50%; of global research output goes to the “Big Five.”
Kennzahl: 19 bn; annual revenue of the sector (2020, in $) — more profitable than the music and film industries.

This is not fraud in the criminal sense — but the same logic: a public good (knowledge) is diverted into private profit, and access to the truth is artificially made scarce.

The Honest Flip Side

Counter-Argument (Steelman)

A dossier that only accuses builds an enemy image of its own. So here are the strongest objections — unvarnished:

Self-correction. All the findings above are science catching itself — replication projects, Retraction Watch, PubPeer, sleuths like Elisabeth Bik. Brian Nosek, who led the large psychology replication, explicitly called his project “a demonstration of self-correction.”
Not exclusively capitalist. See the state half above: Lysenko, “Deutsche Physik,” GDR doping, plan statistics. “Publish or perish” and paper mills are bureaucratic pressure, not the pure market — the highest retraction rates are found in, among others, Saudi Arabia, Pakistan, Russia and China.
Capital funds real science. Without market-driven research there would be neither the scaling of mRNA nor a large part of applied innovation. Profit is not only the enemy of truth, often it is also its financier.

Evidence

Sources

01 · Open Science Collaboration, “Estimating the reproducibility of psychological science”, Science 2015. science.org ↗
02 · Begley & Ellis, “Raise standards for preclinical cancer research”, Nature 483, 2012 (Amgen 6/53). nature.com ↗
03 · Prinz et al. (Bayer), “Believe it or not”, Nature Reviews Drug Discovery 2011. — Baker, “1,500 scientists lift the lid on reproducibility”, Nature 2016.
04 · Lundh et al., “Industry sponsorship and research outcome”, Cochrane Database 2017. cochranelibrary.com ↗
05 · Turner et al., “Selective Publication of Antidepressant Trials”, NEJM 358, 2008. nejm.org ↗
06 · Kearns, Schmidt & Glantz, “Sugar Industry and Coronary Heart Disease Research”, JAMA Internal Medicine 2016.
07 · Oreskes & Conway, “Merchants of Doubt” (2010); Michaels, “Doubt Is Their Product” (2008). “Doubt is our product”: R.J. Reynolds memo, 1969.
08 · Purdue / OxyContin: U.S. DOJ proceedings 2007 & 2020; Porter & Jick, NEJM 1980 (letter to the editor). promarket.org ↗
09 · Vioxx: Graham et al., Lancet 2005 (88,000–140,000 cases); Union of Concerned Scientists, “Merck Manipulated the Science about Vioxx.”
10 · Wakefield: Brian Deer, BMJ 2011; Lancet retraction 2010. Overview ↗
11 · Macchiarini: Science/AAAS 2023 (conviction, 30 months); Lancet retractions 2018/2023.
12 · Theranos: U.S. DOJ, San José 2022 (135 months; $452 million restitution); appeal upheld 2025.
13 · Tessier-Lavigne: Science 2023; Stanford Board of Trustees, Filip report 2023.
14 · Retractions: Van Noorden, “More than 10,000 research papers were retracted in 2023”, Nature 624, 2023.
15 · Publishers: The Conversation 2025; The Nation 2023; Buranyi, The Guardian 2017 (“triple-pay”).
16 · Brandolini's law: A. Brandolini, 2013 (stated publicly). Overview ↗
17 · Vosoughi, Roy & Aral, “The spread of true and false news online”, Science 359, 2018. science.org ↗
18 · Lysenko / Vavilov: deJong-Lambert & Krementsov, Genetics (OUP) 2019 — the 1948 academy session. academic.oup.com ↗
19 · Four Pests campaign / sparrows: Frank et al. (NBER), summarized at UChicago/EPIC; Great Famine death tolls 15–55 million. epic.uchicago.edu ↗
20 · “Deutsche Physik”: Scientific American 2024, “How 2 Pro-Nazi Nobelists Attacked Einstein's ‘Jewish Science’.” Overview ↗
21 · GDR doping (State Plan Theme 14.25): Franke & Berendonk, Clinical Chemistry 1997; Stasi files from 1993; ~9,000 athletes.
22 · Pripiski / Uzbek cotton affair: Cucciolla 2017; Gale, “A Window on the Soviet Breakup” (58,000 affected, ~20,000 charged).
23 · Retraction rates by country: Van Noorden, Nature 624, 2023.