The Burden of Disclosure

What it is

Conflict disclosure rests on a simple model. A listener who learns that an adviser is paid by the outcome should discount the advice, and an adviser who knows the interest will be seen should be more careful. Both steps point towards less compliance with biased advice. The burden of disclosure is the observation that the first step holds and the second does not. In the experiments, disclosure consistently lowered trust, which the authors measured with a single item asking whether the adviser prioritised the listener's best interests, and it raised compliance all the same.

The authors call the mechanism a panhandler effect, borrowing from research on reluctant altruism. Disclosure makes two things mutually known: that the adviser personally gains, and that the gain depends on what the listener chooses. A statement meant to say "I gain if you choose this" can be heard as "please choose this because it helps me". A request made in person, with the outcome visible straight away, is hard to refuse, and the extra compliance reflects that reluctance and not a gift freely given.

The authors tested this reading against rivals. Fairness, a wish to reward honesty and warm glow all predict listeners who feel good about complying. The listeners in these studies were less pleased with their choice after disclosure, liked the adviser less, and reported more discomfort in turning the advice down. Trust fell by similar amounts whichever way the disclosure arrived, yet compliance varied with who made it, whether it was common knowledge and whether the decision was public. That pattern is hard to square with a purely informational account, since the information was the same.

In effect

The first experiment recruited 240 people in Pittsburgh, 120 advisers and 120 choosers, and used two die-roll gambles. Almost everyone in a pretest preferred gamble A, which offered better prizes. In the conflicted conditions the adviser received a die-roll only if the chooser took the inferior gamble B, and the adviser either wrote out that interest or did not. Among choosers whose adviser recommended B, 53 per cent took it without disclosure and 81 per cent took it with disclosure. Agreement that the adviser had prioritised the chooser's best interests fell, from 3.35 to 2.86 on a five-point scale in the restricted sample. Choosers in the disclosure condition reported more wish to help the adviser and more discomfort in rejecting the advice, and a combined pressure measure carried part of the effect on choice.

Five further experiments varied the setting. Raising the stakes in a second study, with a gap of 30 dollars between the gambles, did not remove the effect: choosers took B 30 per cent of the time without disclosure and 75 per cent with it. A third compared disclosure made by the adviser with disclosure made in the chooser's instructions. Among choosers advised to take B, 52 per cent complied when the disclosure was external and 84 per cent when the adviser made it, and the difference in pressure accounted for the difference in choice.

The fourth experiment added common knowledge. Compliance among those advised to take B was 47 per cent when the disclosure was external and the adviser did not know the chooser had seen it, 72 per cent when it was external and mutually known, and 92 per cent when the adviser made it. The fifth let some choosers decide in private. Privately, compliance was 15 per cent without disclosure and 59 per cent with it. Where choosers first decided in public and could then change their minds in private, 88 per cent of those who had heard a disclosure chose B publicly and 43 per cent of them changed their minds. The final rates, 50 per cent against 28, did not differ significantly. The sixth experiment gave advisers knowledge that choosers lacked, and the effect persisted, with compliance of 64 per cent without disclosure and 87 per cent with it, although the mediation test for pressure was not significant in that study.

The same logic has been tested in medicine. A 2016 study by Sah, Fagerlin and Ubel used recorded surgeon and patient consultations at Veterans Affairs hospitals and a randomised laboratory experiment. Disclosing a specialty bias raised patients' trust and raised their likelihood of choosing the treatment that matched the physician's specialty.

What it does not say

It does not say that disclosure never helps. In a separate set of experiments, Sah and Loewenstein (2014) found that when advisers could decline a conflicted payment, 63 per cent accepted it without disclosure and 33 per cent with it, so the label can work on the adviser even where it works poorly on the listener. It also does not say that listeners are credulous. Trust fell in every one of these studies, and a randomised study of internists found that an industry funding statement lowered ratings of a trial's rigour and willingness to prescribe. Reading a printed label in private is a different act from facing the person it describes.

It is also distinct from moral licensing. Cain, Loewenstein and Moore (2005) proposed that disclosure can leave advisers feeling licensed to exaggerate further. That is a claim about the adviser. The burden of disclosure is a claim about the listener, and the two can operate together.

The authors state the limits of their designs. The studies are stylised, the stakes are small compared with many real settings, and the advice in the die-roll tasks was clearly bad, whereas real advice is often ambiguous. Advisers' gains were a die-roll or a gift certificate. Some measures were weak: the pressure composite in the first experiment had a reliability of .59.

The practical conclusions the authors draw are narrow. Disclosure should come from an outside source and not from the adviser. Listeners should not make important decisions in front of the adviser, should have time to reconsider, and should have an impersonal way to decline. Policy, they argue, should focus less on disclosing conflicts and more on decreasing them. Sah's later work on disclosure policy carries the same direction in its title, shifting policies from revealing to resolving conflicts of interest.

---