Risk Disclosure: Writing a Warning People Actually Read

Disclosure salience is whether a warning is actually noticed and understood, as distinct from whether it is present. Presence is a property of the document and salience is a property of the reading, so a complete and accurate disclosure nobody registers has discharged a formality and changed nothing about the decision it was meant to inform.

One sentence ends a great many arguments about whether somebody was warned: it was in the document. The sentence settles presence and leaves salience untouched. A warning exists in two places, and only one of them is the paper. The first place is the document, where the warning can be checked word by word. The second place is the reading, where the warning either registered or it did not. A warning that never reached a reader has produced a document and not a communication, so the second place is the one that decides the outcome.

Fair presentation runs through three routes, as communication conduct sets out: what is said, what is left out, and how prominently each sits. Prominence is the third route, and prominence decides whether the other two ever get read at all. Prominence is also the route that is easiest to satisfy on paper and hardest to satisfy in fact.

What is Disclosure Salience, and why does prominence decide the outcome?

Disclosure salienceWhether a warning is actually noticed and understood, not merely present. is a property of the meeting between a document and a reader. Presence can be audited alone, at a desk, with the document to hand and nobody else in the room. Salience cannot. Half of salience lives in somebody the writer has never met, reading at a speed the writer did not choose, on a day nobody knows anything about. Presence is checkable by a single person and salience is not, so the checkable one quietly becomes the standard.

The shape is not confined to finance. A hospital discharge sheet carries the instruction that matters on the reverse of its third printed side, under a heading that looks like every other heading. A rental agreement states the notice period in the same nine point type as the clause about repainting. A packet of medicine prints the interaction that could send somebody back to hospital in the smallest text on the leaflet, folded twice. In each case the information is complete, accurate and present, and in each case the person who needed it walked out without it. Nobody lied. Nobody left anything out. The failure was entirely a failure of prominence, and prominence is a design decision somebody made.

Because the left column is the one a writer controls completely, it is also the one a writer optimises. Every incentive points that way. A risk paragraph that is longer is more defensible, a caveat that covers one more case is safer, and an extra qualifying clause costs the writer nothing to add. Each of those moves improves the left column and quietly damages the right one, and nobody in the process is doing anything wrong. The failure is durable because careful people following sensible instincts produce it.

ProminenceHow likely a passage is to be read, decided by placement, length and contrast. is the lever that connects the two columns. Prominence is not a synonym for shouting. Prominence is the collection of decisions about where a passage sits, how long it is, what it is set in, and when it arrives, and each of those decisions changes the probability that a particular reader reaches a particular sentence. Salience is the outcome; prominence is the set of choices that produce it. A writer cannot control whether a reader understands, but a writer can control every one of the inputs that decide whether they get the chance to.

What does a warning that is present and unread actually look like?

Meera Sundaram, a salaried investor of 41 built for teaching, decides on her own account through Palash Advisory Services Private Limited, where Devika Rao is the adviser. The Palash decision log records 240 decisions taken by 60 investors across eight quarters, Meera among them.

One scheme document in that record runs to 46 printed sides. The risk statement sits on side 31, set in eight point type. The statement is accurate, complete and delivered. Thirty readers were asked afterwards to state the main risk in their own words. Just 7 of them could, which is 23.3 per cent. The same risk was then written out in 90 words and placed at the top of the first side, and 24 of the 30 could state it, which is 80.0 per cent. The risk did not change. The accuracy did not change. The reader did not change. Only the placement and the length changed, and the share of readers who came away knowing the main risk went from under a quarter to four in five.

Nothing about that placement is dishonest. Side 31 is where the risk section naturally falls once the document is ordered the way documents of that kind are ordered. Eight point type is what fits when the section has to sit beside everything else. Every individual decision that put the statement there was reasonable. The result is a warning that 3 readers in 4 walked past.

Beshears and others, working on simplified fund disclosure in 2011, found the same shape in a real setting: shortening and simplifying what a person is given changes what they take away from it. The measurement in the Palash record was built for teaching, and one small illustration is never evidence that a rule works. One illustration does show the shape of the effect, and the shape is unambiguous. Two documents, both accurate, both complete, both delivered. One communicated and one did not.

Why does length work against a warning rather than for it?

Ask a writer to make a warning safer and the reliable instinct is to make it longer. The extra sentence is true and the reader can simply skip it if it does not apply, so another sentence covering another case cannot possibly make things worse. The reasoning is airtight on the left column above and false on the right one, and what defeats it is arithmetic rather than psychology.

A reader arrives with a fixed budget. Not a stated budget, not a considered one, but a real one: the amount of close attention they are going to spend on this document before they move on. Suppose, purely as an illustration and with no measurement behind it, that a particular reader will read three paragraphs closely and skim the rest. If the warning is one paragraph in a document of three, it gets read. If it is one paragraph in a document of twelve, the chance it is one of the three is a quarter. If it is one paragraph in a document of 46, the chance is under a fifteenth. The words did not weaken. The probability of reaching them did, and the writer produced that fall by adding material.

The curve falls fastest at the start. Most people find that steepness surprising. Going from three paragraphs to twelve costs three quarters of the probability. Going from twelve to 46 costs most of what is left, but there was not much left to lose. The damage is done early, by the first few additions that felt entirely harmless.

None of this is an argument for leaving things out. Completeness and salience are both real properties and a document has to carry both, so the writer is managing a genuine trade rather than choosing between a good option and a bad one. The trade rules out one comfortable belief, that more is free. More is not free. Every addition is paid for by every sentence already in the document, and the risk statement pays the same rate as everything else.

How does the unit a risk is expressed in change whether it registers?

Take one fact out of the invented record and hold it completely still. Meera Sundaram carries Rs 1,80,000/- of card borrowing at 36.0 per cent a year, and her monthly outgo is Rs 55,000/-. The two numbers do not move again. The unitThe form a quantity is expressed in, which changes whether it registers. the cost is stated in does move, and four ways of stating that cost are all exactly true.

Before going any further, the honest part, and it belongs here rather than three paragraphs later. All four statements are equally true, and choosing between them is a choice about how a fact will land. Communication conduct establishes that there is no neutral frame, and there is no neutral unit either, so a writer who picks the most striking of the four has not escaped the choice by picking the most effective one. The unit is a decision the writer makes on the reader's behalf, and the only defence of it is that it serves the reader's understanding rather than the writer's purpose. Everything below describes how the four differ. None of it says the fourth is the right one.

Now the part that makes the unit matter. A rate is not a quantity, so a percentage rate is the least demanding thing to write and the most demanding thing to read. Thirty six per cent of nothing is nothing. Before the number can mean anything at all, the reader has to hold the balance in mind, multiply, and then place the answer against something they care about. Three operations, performed silently, by somebody who is reading a document rather than sitting an examination. Most readers do not perform them, and the statement therefore passes through without leaving a quantity behind.

Bertrand and Morse, in Information Disclosure, Cognitive Biases, and Payday Borrowing in the Journal of Finance in 2011, tested exactly this move in a real borrowing setting. Expressing the cost of borrowing in money rather than as a rate changed what people did. The finding is why the unit question is central rather than a matter of style. The unit is not presentation. The unit is part of what was communicated.

What happens to one fact when the unit it is stated in changes?

Set out in a list the four statements read flat, and flatness is exactly what a list does to them. The control below moves one thing and one thing only: the unit. Behind it, Rs 1,80,000/- at 36.0 per cent a year sits completely still. The statement changes, the arithmetic the reader is left holding changes, and the borrowing does not move at all.

The four also stand in static form, and here they are: 36.0 per cent a year; Rs 64,800/- a year, being Rs 1,80,000/- multiplied by 0.36; Rs 5,400/- a month, being Rs 64,800/- divided by 12, which divides exactly with no remainder; and 9.8 per cent of monthly outgo, being Rs 5,400/- on Rs 55,000/-, which is 9.818 per cent before rounding. Rs 64,800/- on Rs 6,60,000/- of yearly outgo is the identical 9.8 per cent, so the same share appears in the annual form too. All four are equally true and choosing among them is a framing choice.

Where does a warning have to sit, and when does it have to arrive?

Placement sounds like one question and is really two, and confusing them is how careful people produce documents that fail. The first question is where the passage sits inside the thing being read. The second is where it sits in time, relative to the moment the reader stopped being open to changing their mind. Both are called placement, both are decided by the writer, and only the first one ever gets audited.

The document half is the easier of the two. A reader does not begin at the beginning and proceed evenly to the end. A reader meets a first screen, or a first printed side, and what sits there is read at a level of attention nothing later in the document will get again. Everything after that competes with a reader who has already started deciding whether this is worth their time. So the practical question is not where the risk statement sits in the running order of the sections; it is whether it is inside the part the reader was always going to look at.

TimingWhen a warning arrives relative to the decision it concerns. is the harder half. Every decision has a moment before which it is genuinely open and after which it is effectively closed. The closing moment is rarely the signature. The closing moment is usually earlier, in a conversation, when the person made up their mind and everything afterwards became administration. A warning delivered after that moment has not been read badly. The warning has been read at a point where reading it changes nothing, and that is a different failure and a worse one.

Take 12 October in the invented record. Meera Sundaram sells Suvarna Chemicals Limited whole at Rs 4,60,000/-, booking Rs 60,000/- on a cost of Rs 4,00,000/-, and keeps Kesari Logistics Limited, then showing a loss of Rs 1,05,000/- against its Rs 3,00,000/- cost, saying she will sell it when it returns to what she paid. Suppose a warning about holding a position because of its purchase price had been in her hands. If it reached her while she was still weighing the two, it had somewhere to land. If it reached her in the confirmation afterwards, it was a sentence about a decision that was already made. The identical words are a warning in one position and a receipt in the other, and only the clock decides which.

Timing is the hardest of the four levers to satisfy honestly. Position in a document can be checked by looking at the document. Length can be counted. The unit can be read off the document. But whether a warning arrived while the reader was still open requires knowing something about the reader that no file records. Arrival time therefore sits in the right hand column above, a property of the reading and not of the paper.

How to Review Risk Disclosure for Salience: what is asked, and in what order?

Everything so far has been description. A review that is not in a fixed order becomes an impression, and an impression is exactly what a long careful document is best at defeating, so the description now turns into five questions asked in a fixed order. The order below is not sacred, but having an order is, and the reason is that each question is easy to answer honestly on its own and very easy to skip once a view about the document as a whole has formed.

Question one asks where it sits, in two senses at once: where in the running order of sections, and whether any part of it appears in the first thing the reader is shown. A risk section that begins on side 31 answers the first sense reasonably and the second sense with a flat no. The earlier arithmetic applies to the whole document and not to the risk section alone, so question two asks how long the surrounding document is. Forty six printed sides is not a criticism of the risk statement. Forty six sides is a statement about the odds facing every passage in the document, and the risk statement is one of them.

Question three asks what unit the risk is expressed in, and how many silent operations stand between that unit and a quantity the reader can place beside something they already track. Question four asks when it arrives, and that question has the fewest documentary answers and the largest consequences. Question five asks whether a reader who has been through it can say the main risk back in their own words, and it is the only one of the five that produces evidence rather than a forecast. The first four questions predict and the fifth measures. The first four are therefore done routinely and the fifth almost never is.

Run the five over the two versions of the invented scheme document and the difference stops being a matter of taste. The 46 side version answers question one with side 31, question two with 46 sides, question three with a description carrying no quantity, question four with the pack handed over at signature, and question five with 7 of 30, or 23.3 per cent. The 90 word version answers with the top of side 1, ninety words, a form the reader can place, before the conversation, and 24 of 30, or 80.0 per cent. Four of the five answers were changed by the writer, and the fifth changed on its own.

Two honest limits belong here. The first is that the review says nothing about whether the disclosure is right. A beautifully placed, short, well united, early warning that is wrong is worse than a buried one that is correct, and accuracy is a separate review that has to happen anyway. The second is that questions one to four are proxies. The proxies are good, they are cheap, and they can be run on a document nobody has read yet, and that is their whole value. But they forecast the fifth answer rather than replacing it, and a writer who never once asks a real reader to say the risk back has never tested the only claim that matters.

Can disclosing a conflict make things worse?

Everything up to this point has been about making a warning land harder, and the natural conclusion is that a warning which lands is always an improvement on one that does not. For most warnings that holds. For one particular kind it does not, and the exception is documented well enough that leaving it out would be comfortable and wrong.

The kind is a warning about the person giving the advice. Not a warning about a holding, or a market, or a risk in the world, but a sentence to the effect that the adviser is paid more if the client takes this. The expectation is straightforward and almost everybody shares it. The listener hears the sentence, marks the advice down accordingly, and ends up better placed than a listener who never heard it. Nothing about that chain looks doubtful when it is written out.

Notice what that sentence does to the room before anybody has acted on it at all.

Loewenstein, Cain and Sah reported in The Limits of Transparency, in the American Economic Review in 2011, that disclosing a conflict of interest can leave the person advised worse off than no disclosure at all. The earlier experimental work by Cain, Loewenstein and Moore, in the Journal of Legal Studies in 2005, set out the same uncomfortable direction. The finding is not that disclosure sometimes fails to help; it is that it can actively hurt the person it was meant to protect. That is a much stronger statement, and the effect has a name: this is the backfireA disclosure leaving the person advised worse off than none would have. result, and it is why this section belongs in a treatment otherwise given over to making warnings land.

What are the two mechanisms behind the backfire, and does either need the other?

Two separate things produce the result, they sit on opposite sides of the conversation, and each of them is enough on its own. A reader who fuses the two into one story will conclude that fixing either half fixes the problem, and that is not what the structure says, so keeping the two apart matters.

The first is moral licensingFeeling absolved by having disclosed, and acting worse as a result., and it sits on the adviser's side. Having said the difficult sentence out loud, the adviser feels the difficult part is done. Responsibility has been handed across the table: the listener now knows, so whatever the listener does with the knowledge is the listener's affair. The natural consequence is that the advice given afterwards leans further in the adviser's own direction than it would have if the sentence had never been said. Nobody experiences this as dishonesty. Licensing feels like having been straight, and then being free.

The second is social pressureFeeling unable to discount advice after the adviser has been candid., and it sits on the listener's side. The adviser has just been candid about their own interest, at a cost to themselves, in a room where the two of them have to keep talking. Marking that advice down now is not a private arithmetic operation. Marking it down is a visible act with a meaning: the listener thinks a person who has just been honest is going to mislead them anyway. Most people will not do that, and the result is that they follow the advice more closely rather than less. Everyday life is full of the same shape. A shopkeeper who says frankly that the cheaper item is his own brand has made it socially harder, not easier, for the customer to walk out with the other one.

The grid argues something easy to skim past, so read it slowly. If only one cell were shaded, the two mechanisms would be a pair of conditions that had to coincide, and a practice could break the pairing. Three cells are shaded. Each mechanism is independently sufficient, so removing one of them relocates the problem rather than solving it, and the two together simply compound what either would have done alone. An adviser trained out of feeling absolved still faces a listener who cannot comfortably discount. A listener coached to discount freely still faces an adviser who has handed over responsibility.

What does the backfire result not say about disclosure?

The wrong inference is shorter and therefore travels further, so the two are worth setting beside each other in plain words. The refused inference and the supported one begin from the identical evidence and end in opposite places, and the difference between them is whether disclosure is being treated as a cure or as a message.

There is an ordering hiding inside the supported conclusion, and it is about conduct rather than about anybody's money. Where a conflict can be removed, removing it settles what a disclosure only raises. Where it genuinely cannot be, the disclosure carries all the weight, and everything set out earlier about position, length, unit and arrival time becomes the difference between a sentence that was said and a sentence that was received. A warning that merely gets spoken now has a documented way of making things worse, so the backfire result raises the standard a disclosure has to meet rather than lowering its value.

What follows for anybody writing a warning?

Four things follow, and the first three are craft. Put the risk where the reader already is rather than where the section order puts it. Keep the surrounding document short enough that the risk is not competing with forty five printed sides of true and irrelevant material. Express the quantity in a form that arrives finished, with nothing left to be multiplied silently before it means anything. Deliver it before the reader has closed the question rather than with the paperwork that follows.

The fourth is not craft, and it is the one that will not go away. There is no neutral unit. Rs 5,400/- a month is not the true version of 36.0 per cent a year; it is another true version, chosen because of what it does to a reader. A writer who reaches for the most striking of four equally accurate statements has made a choice on somebody else's behalf, and the only defence of that choice is the one they would be willing to say out loud to the reader. If the reason the unit was picked cannot be stated to the person reading it without embarrassment, it was picked for the writer.

Step four is doing more work than it looks. A reader given Rs 5,400/- a month and nothing else has been handed a conclusion. A reader given Rs 5,400/- a month together with Rs 1,80,000/- at 36.0 per cent, the multiplication and the division by 12, has been handed the conclusion and the route to every other form of it. The second reader can check the writer. Showing the working is the difference between choosing a unit to make a fact receivable and choosing one to make it persuasive, and the difference is visible in the document rather than taken on trust.

Set on one frame, the whole subject comes down to four levers and one refusal.

Meera Sundaram, Devika Rao, Palash Advisory Services Private Limited, the Palash decision log, Suvarna Chemicals Limited and Kesari Logistics Limited are invented.
Educational material. Not advice on any investment, tax, budget or market position.