Narrative Economics: What a Story Adds to a Price

Narrative economics studies how accounts of the world spread through a population and what they do once enough people hold them. A story is not decoration on a decision. A story is what travels, and it travels on its own properties rather than on whether it is true. The premium it adds is measurable, and it decays.

All of it rests on taking contagionSpreading from person to person at some rate, with some rate of stopping. seriously as a shape rather than as a figure of speech. An account has a rate at which it passes between people and a rate at which people stop bothering to repeat it. The two rates govern how far an account gets and how long it lasts, and neither of them has any input for whether the account is correct. Everything else follows from those two rates, including the uncomfortable part about decay.

What is narrative economics, and what is it not?

Robert Shiller set the field out in Narrative Economics, published in the American Economic Review in 2017. The proposal was blunter than it sounds. Economics already had events, and it already had measurements of what people did. The missing object was the account itself, counted the way an event is counted: the sentence people say to each other about why something is happening, treated as a thing with its own life, its own rate of travel and its own death.

The clearest starting point lies away from money entirely. A single school in a town gets a reputation for being strict. The origin is usually one incident several years old, and almost nobody repeating the reputation was present for it. The incident is one object. The sentence "that school is strict" is a different object, and it has properties the incident does not have: it is four words long, it fits into a conversation at a bus stop, and somebody can pass it on without knowing a single fact about the school. Narrative economics is the study of the second object.

Narrative economics is not the claim that people are moved by stories, a claim neither new nor measurable. Everybody already agrees that stories move people. The move this field makes is to stop treating the story as commentary attached to a decision and start treating it as the unit that spreads, with a count attached: how many people hold it now, how many held it last quarter, and what the rate of change between those two numbers is. Once that count exists there is a measurement, and a measurement is what separates a field from an observation.

The three counts are the whole measurement, and they are unglamorous. In the invented spread used from here on, four of the sixty people were repeating the account at step seven and two of them at step eight, a fall of 50.0 per cent in a single step. A reading looks like nothing more than that.

What two rates govern how far an account gets?

The shape is borrowed openly. Epidemiologists describe something spreading with two numbers: how often a person carrying it passes it to somebody who has not got it yet, and how often a person carrying it stops carrying it. An account behaves the same way. Somebody repeats it at dinner and two people who had not heard it now hold it. Somebody else who has been repeating it for three months gets bored and stops. Passing on and stopping run against each other, and the contest between them gives the whole life of an account, from the first telling to the last.

Three regimes fall out of the arithmetic, and only three. Where passing on runs faster than stopping, the count of people repeating it climbs. Where the two run level, the count sits still. Where stopping runs faster, the count falls. Whether the account is correct is not an input to either rate, so an account can climb through all three regimes without a single person ever checking anything.

Here is the part that catches people out. The passing-on rate depends on how many people are still susceptibleAble to pick something up because they have not encountered it yet. Once nearly everybody has heard an account, there is almost nobody left to pass it to., and that pool shrinks as the account spreads. Early on, almost everybody a repeater meets is new to it. Late on, almost everybody has heard it already, so the same amount of talking produces almost no new holders. Meanwhile the stopping rate keeps working through the people who already hold it. The count therefore turns over and starts falling on its own, with nobody arguing against anything.

Count the same sixty people three times and the drain becomes visible. After the first step, 12 are repeating it, 3 have already dropped it and 45 have not heard it at all. After the third, 26 are repeating it, 19 have dropped it and only 15 are left to reach. After the eighth, 2 are still repeating it, 57 have dropped it and 1 never heard it. The pool of people who could still be told is what runs out first, and its emptying is what turns the curve over.

Work it through on an invented spread among the 60 people in the Palash decision log. Six of them hold the account at the start. In the first step nine more pick it up and three drop it, so twelve hold it. In the next step fourteen pick it up and six drop it, so twenty hold it. By the third step it reaches twenty six, and that is the top. From the fourth step onward the pool of people who have not heard it has almost run dry, so more people drop it in each step than pick it up. Fifty nine of the sixty ever repeat it at some point; fifty seven of them stop; two are still repeating it at the end.

What makes one account travel further than another?

Four properties do most of the work, and all four can be tested on any sentence heard three times this month. The account is short, short enough to survive being repeated by somebody with thirty seconds. A person is easier to hold in the head than a rate, so the account has a person in it rather than only a quantity. The account explains something the listener had already noticed and had no explanation for, so it lands as recognition rather than as information. And it has repeatabilityHow easily something can be passed on by somebody who does not understand it.: it can be handed on by a person who could not defend a word of it.

Repeatability decides most contests, and repeatability is worth sitting with. A sentence that has to be understood before it can be repeated has a small pool of possible carriers. A sentence that can be repeated by anybody has the whole population. Take a careful account of why a business is doing well, needing three qualifications to survive a retelling, and set it against a four word version that needs none. The careful one loses, and it loses on transmission rather than on merit.

The second list decides whether an account is right. A correct account rests on evidence anybody can go and check, says what observation would show it wrong, has survived a test that could have failed, and holds when the person telling it is replaced. No item on the travelling list implies any item on the truth list, and no item on the truth list helps an account travel. The two lists are independent of each other, and no amount of one buys any of the other.

The everyday version is a street corner. A food stall gets a reputation for the best breakfast on the road, and the reputation reaches four streets away. The reputation travelled as five words, an owner with a name, and a claim every listener could nod at. Eating the breakfast settles whether it is any good, and almost nobody four streets away has eaten it. The reputation and the breakfast are separate objects, and only one of them has salienceThe quality of standing out enough to be noticed and remembered rather than passed over. at that distance.

The invented practice record has a measurement of exactly this. The scheme document runs to 46 printed sides, with the statement of the main risk on the 31st in eight point type. Asked afterwards, 7 of 30 readers could state that risk, or 23.3 per cent. Shown the same statement compressed to 90 words and placed at the top, 24 of 30 could state it, or 80.0 per cent. Same content, same readers, same room. Only the ease of carrying the sentence out of the room changed, and that alone moved the count by more than three times.

Watch what a chain of retellings does to a sentence that started at 27 words. The sentence does not get corrected on the way and nobody argues with it. The sentence gets shorter, and by the time it is down to 6 words it is finally light enough to go everywhere.

What is a Narrative Premium, and how is it measured?

A narrative premium is the part of a level attributable to a widely held account rather than to the cash the holdings actually produce. The premium is one subtraction. Take the level as it reads. Take the level the cash flows on their own would support. The difference is the account's contribution, and it can be written down at any moment the account is being held.

Work it on the invented Palash 100 index, an illustration throughout and never a description of any real market. The index opened at 100.0. Its highest quarter end reading is 131.0, at the Q2 quarter end. Now suppose the cash the underlying holdings produced supports a level of 113.0. The 113.0 line is made up too, and it serves as a fixed reference rather than as a valuation of anything. The premium at the peak is 131.0 less 113.0, or 18.0 points. Expressed against the line, 18.0 over 113.0 is 15.9 per cent. 18.0 points on its own is a distance and 15.9 per cent is the size of that distance against the line it sits on, so both halves of the measurement matter.

The subtraction is not reserved for the peak. Done at every quarter end it gives a reading each time: 5.0 points above the line at Q1, 18.0 above at Q2, 1.0 point below at Q3, 9.0 below at Q4, and then 3.0, 11.0, 8.0 and 14.0 above it again. The premium is a quantity with a value at every single reading, and the peak is simply the reading where that value happened to be largest.

Why does the premium go without anybody refuting the account?

Follow the same invented index past its peak. The Q2 quarter end reads 131.0. The Q3 quarter end reads 112.0, already 1.0 point under the 113.0 line. The Q4 quarter end reads 104.0, a full 9.0 points under it. Measured from the peak, that is a fall of 20.6 per cent, and the premium of 18.0 points did not merely disappear. The premium went past zero and out the other side.

Now ask what happened in between, and notice what did not. Nothing on the Palash decision log records anybody producing evidence against the account. There is no entry where the account was tested and failed. The log carries the amount of activity instead: turnover ran at 3.1 times its eight quarter median in the quarter ending at the peak and 0.4 times in the quarter ending at the low. The account was not defeated; it stopped being said, and the premium went with the saying rather than with any argument.

An account fading without ever being refuted is decayAn account being repeated less often, without being refuted. Nobody argues against it; people simply stop bringing it up., and decay is the hardest part of the mechanism to accept. An argument ends when one side is disproved, so every intuition says a premium ought to disappear the same way. Accounts almost never end that way. Accounts end the way the count of repeaters falls in the third regime: quarter after quarter, and nobody says anything at all. The overshoot below the line is that same absence taken one step further. A population that has just stopped repeating something does not return to the state it was in before it ever heard it.

The fall splits into exactly two named pieces, and naming them is what stops it being one vague slump. From 131.0, the first 18.0 points take the level back to the invented line at 113.0, and that piece is the premium going. The next 9.0 points take it to 104.0, and that piece is the overshoot. Together they are 27.0 points off 131.0, the fall of 20.6 per cent.

Why does an account outlive the facts that started it?

Because the account and the facts get separated at the very first retelling, and after that they travel apart. Somebody sees something and puts it into a sentence. The person who hears the sentence has the sentence and not the something, and when they pass it on they pass on what they have. By the fourth or fifth telling nobody in the chain has any access to what began it, and the account is still perfectly repeatable. The account has been repeatable all along, and repeatability is the property that got it this far.

Return to the invented spread among the 60 people. Six of them held the account at the start, and fifty nine of them ever repeated it. Fifty three people therefore acquired it entirely from other people, roughly nine in ten of everybody who ever said it. Nothing in the passing on ever consults the observation that started the account, so an account can be true when it starts, false a year later, and completely unchanged in the mouths of the people repeating it. There is no step in the chain at which the original gets rechecked, unless somebody deliberately builds one.

The pattern is an entirely ordinary one. A road has been closed for repairs and the whole neighbourhood knows to take the long way round. The repairs finish. The road opens. Six weeks later people are still taking the long way, and if asked why, they will say the road is closed, with complete confidence and no recent evidence. Nobody lied. Nobody was careless. The account outlived its facts because repeating it never depended on them.

Set the two objects side by side and the asymmetry is the whole point. The observation reached six people and was available for one step. The account reached fifty nine people and was still being said eight steps later, by which time nobody repeating it could have produced the observation even if somebody had asked.

How can an account carrying information be told from one carrying only itself?

One question does it, and it is asked of the people repeating the account rather than of the account. The question is what observation would show this wrong. An account that carries information has an answer: something that could be looked at which, if found, would settle the matter against the account. An account carrying only itself has no answer, and the people repeating it are usually surprised the question was asked at all.

The question is the ordinary test for whether a statement is falsifiableStated so that some observation would show it wrong. A claim nothing could contradict is not making a claim about the world., applied to a shared account rather than to a scientific theory. Passing the test establishes less than it seems to. Passing it does not make the account right. Passing it makes the account the kind of thing that could be checked. A checkable account names what to go and do next, and that is a much lower bar and a far more useful one. An account nobody can attach a defeating observation to is not a weak claim about the world; it is not a claim about the world at all.

The test applies to two versions of the same sentence. A business is doing well because a new product is selling; the defeating observation is the sales, and anybody can go and find them. A business is doing well because it has momentum. Whatever happens gets folded back into the account, so no observation counts against momentum. The first version transmits something. The second transmits itself. The momentum version is shorter and never has to survive a fact, so it will travel further.

What can a person do with a shared account, and what can they not?

Devika Rao, the adviser at the invented Palash Advisory Services Private Limited, does one small thing with all of this and refuses to do a second. The small thing is a column. Beside each decision she records the account being relied on, in one sentence, and beside that the observation that would show the account wrong. Two boxes. Where the second box is empty, the decision is not blocked and nothing is forbidden; it is simply marked as resting on an account that has nothing attached to it.

The second use is a count she takes at the end of each quarter: how many decisions in the book cite the same account. The count is not about the holdings at all. The count says how much of the reasoning across many different people rests on a single sentence, a fact about the reasoning rather than about any market. A person deciding alone, with no adviser and no committee, can take the same count over their own last ten decisions in about five minutes.

The sheet itself is two columns and nothing else, and the empty cells on the right are the reason it exists at all.

The invented log has something to say about whether writing it down changes the writing. Twenty of the sixty investors adopted a written checklist on 4 November. Across the last four quarters they recorded a written reason on 34 of 41 decisions, or 82.9 per cent, against 19 of 63, or 30.2 per cent, for the other forty. The comparison shows that the reasons got recorded and nothing whatever about returns. Sixty people over eight quarters cannot carry a claim about returns, and no such claim is made.

The limit on all of this is the important half. None of this produces a reason to buy anything, sell anything, wait or avoid. A premium measurable at the Q2 peak is a premium measured with the Q2 peak already in view, and the same subtraction done at the time would have needed a cash flow line nobody had. Where a duty to record the basis of a recommendation applies to a registered intermediary, that is a conduct question, and the Securities and Exchange Board of India at sebi.gov.in is where such requirements live.

Where is a shared account doing genuine work?

Concluding that shared accounts are a defect would be a misreading. A shared account is how anything gets from the person who worked it out to everybody else. Nobody rebuilds arithmetic from scratch before using a receipt, and nobody re-derives what a kilogram is before buying vegetables. A population that holds the same account of how something works can coordinate without each member repeating the work, and that saving is enormous and almost entirely invisible.

The useful ones share a property. An account doing genuine work carries a defeating observation along with it, so the saving in effort does not come at the cost of the ability to notice when the account has stopped being right. A weather warning travels fast, is short, has people in it, and can be repeated by anybody, all four travelling properties at once. The warning also names what would show it wrong, and the answer is the sky. Both lists at once. Scoring on both lists at once is rare, and worth recognising when it appears rather than treating every widely held account as suspect.

What does narrative economics leave unexplained?

Rather a lot, and being specific about it is what keeps the subject honest. The travelling properties say what helps rather than what wins, so narrative economics does not say which of several competing accounts will spread. The cash flow line a premium is measured against is exactly what nobody has at the time, so narrative economics does not give the size of a premium before the fact. Nor does it date the turn: the fall from 131.0 to 104.0 has a shape, and the shape does not carry a calendar.

And it does not hand anybody a reading of where a level currently sits against anything. Every measurement above was taken with the whole eight quarters already on the table. A premium identified after the fact is a description of what happened and not a reading anybody had available before it, and the difference between those two is the entire distance between an explanation and an instruction. The Palash 100 index and the 113.0 line are made up, and the series ends at Q8 because the invented log records nothing after it.

Meera Sundaram, Devika Rao, Palash Advisory Services Private Limited, the Palash decision log, the Palash 100 index and Suvarna Chemicals Limited are invented.
Educational material. Not advice on any investment, tax, budget or market position.