Signup terms, conflict-of-interest lines, and the EULA under every install button all rest on the same legal bet: warn someone, and they will read the warning and change course. Clickstream data from real shoppers puts that bet near zero.
Unread text is only the first failure. Once a conflict of interest is named out loud, the advisor can lean harder into the bias the disclosure was meant to contain, and the listener still fails to discount far enough. The safeguard itself can amplify the skew it was meant to contain.
Both halves land here: how rarely standard disclosure gets opened, why naming a conflict can backfire when someone does read it, and which cost formats have actually shifted borrowing in the field.
The story in four parts
Browsing logs across a large set of software checkouts show almost no one opens the linked standard-form contract before purchase.
In a controlled lab study, disclosing a real financial conflict left advice more skewed than before, and the people hearing it still under-corrected.
Inside real payday stores, showing dollar fees that pile up over renewals reduced future borrowing. APR and credit-card comparison sheets barely moved it.
The evidence points at removing the conflict a disclosure tries to flag, and at showing costs the way that payday field test showed they change behaviour.
Account terms, install licences, and checkout fine print all treat the “I agree” click as proof the document was read. That is a legal fiction. A large study of real online shoppers measured how often the fiction matches behaviour.
For every thousand visitors who land on a page with a EULA, how many actually open it?
Bakos, Marotta-Wurgler, and Trossen followed clickstream logs across 90 online software firms and about 48,154 monthly shoppers. The question was whether anyone opened the End User License Agreement (EULA) linked from checkout, counted by the click itself rather than by what people claimed they would do.
What it teaches: roughly one or two shoppers in a thousand opened the licence before buying. Everyone else clicked agree without seeing the text. Careful wording cannot steer a decision nobody encounters.
That result measures attention. It leaves open what happens on the rare days a disclosure is read. The sharper half of the problem sits on this site's Disclosure Backfire principle. In a lab study, naming a financial conflict of interest left the advice more skewed than before, and listeners still under-discounted the bias.
Unread contracts and backfiring conflict lines both leave the same practical gap. Which way of showing a cost actually changes later behaviour? Bertrand and Morse ran that test in payday lending, with real borrowers and real money on the line.
Dollar fees over time were the format that moved borrowing. An APR and a credit-card comparison were accurate disclosures, and both left later borrowing close to the control path. Showing the cash a borrower would pay in fees under repeated renewals cut subsequent borrowing by roughly 11% over four months.
Dollar-fees-over-time worked because it put a concrete, personally relevant figure in front of someone at the decision itself. That is a narrow win for one delivery format, not proof that any disclosure succeeds once the wording is polished. Two limits sit next to that result.
Bertrand and Morse handed the sheet over in person, at the counter, during the loan. A long disclosure tucked into onboarding has neither that timing nor that delivery, and Bakos et al. already show most people never open the linked page no matter how it is written.
A clearer conflict disclosure also leaves the incentive behind the conflict in place. The Disclosure Backfire principle on this site notes that a fee-only advisor can name their pay without the same backfire, because there is no percentage upside left to lean into. Rewording the same conflict still leaves that upside standing; it only states it more plainly.